Smart glasses incidents at university entrance examinations in Seoul, disclosed on October 13, have exposed a gap between checking candidates before a test and maintaining security throughout it. Devices that combine cameras and connectivity with ordinary-looking frames are increasingly difficult to identify through visual inspection alone. Reports that a candidate changed glasses after an initial check highlight the limits of treating entry screening as sufficient protection. Exam security must cover what happens during the test and how suspected violations are investigated afterward.

Yonsei University identified a candidate wearing smart glasses during its mathematics essay examination for natural sciences and interdisciplinary applicants on October 10. The University of Seoul detected a similar incident during its natural sciences essay examination on October 3. The Ministry of Education plans to include smart glasses guidance in instructions and training videos for College Scholastic Ability Test supervisors and share the materials with universities conducting their own examinations. With the Korea Institute for Curriculum and Evaluation already listing smart glasses among prohibited items for this year’s CSAT, the central issue is how those rules are enforced.

Cameras and Connectivity Matter Beyond AI Accuracy

Smart glasses differ in their capabilities and operation. Some combine cameras and speakers, while others include an in-lens display for messages and other information. According to Meta’s official materials, Ray-Ban Meta glasses can use a camera to answer questions about what the wearer sees, while its separate display-equipped product can show AI responses and messages. Exam supervisors therefore need to understand a range of functions rather than assume that all smart glasses work in the same way.

AI processing does not necessarily take place inside the glasses. Ray-Ban’s official guidance explains that some visual AI functions send a photograph to Meta’s cloud for processing before delivering an audio response. Its documentation also describes a setup involving a smartphone, an app and internet access. This architecture extends the potential security concern to connected devices and external services. The products, connection methods and actual receipt of AI responses in the Korean incidents require separate verification.

The risk consequently extends beyond whether AI can solve an examination question correctly. Photographing questions or transmitting them outside the room can itself threaten exam security. An inaccurate AI response does not eliminate the possibility of unauthorized information exchange. Rules and inspection procedures should address photography, storage, communication and the receipt of outside information alongside AI capabilities.

Security Must Continue After the Initial Check

The reported change of glasses after inspection at Yonsei points to an operational issue separate from identifying a device. Checking ordinary glasses before the examination provides limited protection if a candidate subsequently puts on another pair. Supervisor training needs to cover how changes in worn devices are identified and handled. Pre-exam screening and supervision during the test should form a continuous process.

Universities could establish clear procedures for declaring and storing spare glasses or assistive devices, as well as for replacing them during an examination. Candidates who need a replacement could be required to obtain a supervisor’s approval, with uncertain devices referred to the examination administration team. Such arrangements would help reduce differences in treatment arising from individual supervisors’ experience. Candidates should receive the procedures in advance to minimize confusion in the examination room.

Identifying smart glasses primarily by thick frames or a particular design is unlikely to remain effective. It can place unnecessary suspicion on candidates wearing conventional glasses, while changes in product design can weaken existing identification guidance. Training should cover functional features such as cameras, controls and charging contacts, supported by a process for examining uncertain devices. An unusual appearance may justify a closer check, but it should remain distinct from evidence of misconduct.

Protecting Questions as Well as Answers

Camera-equipped devices create risks beyond the integrity of an individual answer sheet. Examination questions, nearby candidates’ answers and the room itself may also be photographed. Where the same questions are used in sessions held at different times, unauthorized disclosure could affect the fairness of later examinations. Institutions therefore need both a process for determining individual rule violations and a way to assess any broader security impact.

Investigations should distinguish possession or wearing of a device from its operation, photography and external transmission. Consistent records of when a device was discovered, its condition and the supervisor’s observations can support subsequent decisions and appeals. Claims about receiving outside information or using it to produce answers should remain within what the evidence establishes. Accurate fact-finding strengthens confidence in both disciplinary action and examination administration.

Inspection procedures must also account for vision correction and disability support. Advance guidance for candidates using ordinary glasses or approved assistive devices can reduce anxiety and delays. Institutions need to apply prohibited-device rules consistently while protecting candidates’ ability to take the examination. Fairness depends on both preventing misconduct and reducing mistaken suspicion.

Australia Uses Oral and Supervised Tasks to Verify Understanding

Australia’s Tertiary Education Quality and Standards Agency, or TEQSA, has outlined assessment approaches that require students to explain or demonstrate what they have learned. Its examples include oral presentations, in-class tasks and supervised practical demonstrations. One approach incorporates at least one secure assessment in every subject, which students must satisfactorily complete to pass. This gives educators evidence that students can independently understand and apply key material alongside the quality of their submitted work.

In practical terms, the approach can mean asking a student to explain the reasoning behind a report’s conclusion or demonstrate a procedure learned in a practical class. TEQSA also describes supervised assessments at several points throughout a degree rather than relying on a single check near graduation. The document provides guidance for university assessment rather than mandating an identical examination format across Australian institutions. Any application to Korean university admissions would require separate consideration of selection objectives, candidate numbers, marking consistency and operating costs.

Universities could consider placing greater emphasis on reasoning in written answers or using follow-up interview questions about why a candidate reached a conclusion and how changed conditions would affect it. These measures aim to establish whether applicants understand their own answers. Changing questions alone, however, cannot prevent unauthorized assistance through smart glasses. Assessment design needs to develop alongside device checks and effective supervision.

Effectiveness Requires More Than Counting Incidents

The number of detected cases is an incomplete measure of success. An increase may reflect more attempts, better supervision or both. The ministry and universities should also examine supervisor training, the time needed to verify suspicious devices, mistaken identifications, support from examination administrators and responses to possible question leaks. The challenge exposed by smart glasses extends beyond finding electronic devices at the entrance. It requires procedures that maintain security through the end of the examination and handle unusual situations consistently.

Read the Korean version

Series Order

Part 1 | What Should Humans Do When AI Expands the Limits of Thought?
Part 2 | Companies That Adopt AI and Companies That Create with AI Are Different
Part 3 | In the AI Era, Corporate Competitiveness Depends on the Speed of Organizational Change

By Im Byung-min (IBM), Columnist

Most companies today are adopting AI. Generative AI has become a hot topic in meeting rooms, employees receive AI training, AI chatbots are deployed in customer service centers, and coding AI is introduced to development teams. On manufacturing floors, vision AI inspects quality; marketing departments use AI to craft ad copy; and executives demand AI-related strategies and roadmaps. On the surface, it appears that corporate AI transformation is already underway.

However, we need to ask ourselves a critical question. Have we simply adopted AI, or is the company itself changing through AI? These are two entirely different matters. AI adoption is a technological issue, whereas AI transformation is a management issue.

It is not difficult for a company to purchase an AI solution or build a single system. Selecting a consulting firm, implementing a platform, and running a few pilot projects are also feasible tasks. The real challenge lies in what follows. Is AI actually changing work processes, the speed of decision-making, organizational roles, and the way people work? Ultimately, is it transforming the value delivered to customers and the company’s competitive approach?

This is where a vast gap emerges between AI adoption and AI transformation. AI adoption involves layering a new tool—AI—onto existing workflows. AI transformation, by contrast, entails redesigning existing work processes with AI as the fundamental premise.

Consider, for instance, providing a sales representative with generative AI. If the AI summarizes customer meeting details, drafts proposals, and searches for market information, that certainly constitutes AI adoption. However, the situation changes if sales operations—including workflows, evaluation systems, customer management methods, and the way sales data is accumulated—are redesigned with AI at the core. That constitutes AI transformation.

Similarly, a manufacturing company does not automatically become an AI-transformed enterprise simply by installing an AI-based vision inspection system. True transformation begins only when AI goes beyond merely detecting defects to linking production conditions with quality data, automatically adjusting processes, and even redefining the role of the quality management organization. Ultimately, what matters is not merely the presence of AI, but how deeply it is integrated into the company’s operating system.

In the AI era, a company’s competitiveness will likely hinge not on the number of AI projects it has, but on how quickly it can scale AI adoption across the entire organization. Some companies run dozens of AI projects. They generate countless reports, conduct proof-of-concept (PoC) projects, and hold seminars and training sessions. Yet, even after a year, the actual way work gets done on the front lines remains largely unchanged.

In contrast, other companies may start with only a few AI projects but quickly validate them, decisively halting those that fail while immediately scaling successful ones to other parts of the organization. When an individual becomes proficient with AI, their know-how is codified into the organization’s workflows. AI-driven methods developed in one department are promptly adopted by others. This triggers a chain reaction of changes across workflows, systems, data structures, and evaluation frameworks. The difference between these two companies lies not in their level of AI technology, but in the speed of their transformation.

In the AI era, this gap widens over time because AI transformation does not proceed linearly. It evolves from an individual using AI to a team adopting it, eventually changing the team’s working methods and establishing them as the standard for the entire organization. As standards shift, systems evolve and data accumulates. The level of AI utilization then rises, ultimately leading to increased productivity.

Furthermore, companies that transform quickly build a structure that enables even faster future transformation. Conversely, slow-moving companies may find that adopting AI actually increases complexity. Departments use disparate AI tools, data remains siloed, and new AI-driven processes clash with legacy workflows. Despite the adoption of AI, employee workloads increase, leading to the inevitable question: “Why hasn’t productivity improved?” This is not a failure of AI itself, but a failure of the transformation strategy.

When formulating AI strategies, many companies begin by asking what technology to implement. However, the order of questioning must change. What is the first task within our company that needs to change? How can AI redesign that task? And how quickly can we scale that change across the entire organization?

These questions must take precedence. In the AI era, a leader should not necessarily be the person with the most technical knowledge of AI, but rather someone who manages the pace of change. Instead of focusing solely on where to apply AI, leaders must decide which tasks to eliminate, which decision-making structures to alter, and how to redefine organizational roles. Above all, it is crucial to abandon perfectionism. Traditional corporate decision-making typically involves a sequence of careful review, approval, securing a budget, system development, and extensive testing before implementation.

Consequently, the most dangerous state for a company is the illusion that it is adopting AI while its fundamental nature remains unchanged. AI strategies may exist in boardrooms and AI roadmaps in reports, with AI projects underway across the organization. Yet if employees continue to work in the same old ways, decisions are made at the same old pace, and key performance indicators (KPIs) remain unchanged, it is difficult to call this AI transformation.

Read the Korean version

The South Korean power equipment maker announced on October 13 that it had signed a contract worth $138.46 million (approximately KRW 185.3 billion) to supply switchgear and distribution transformers to Bloom Energy.

Under the agreement, LS Electric will provide 38-kilovolt high-voltage switchgear and distribution transformers for an on-site power generation project serving a major technology company’s data center in Pennsylvania. Deliveries are scheduled from April through July 2027.

The contract marks another step in the companies’ expanding partnership, which began in April 2025. The growing order volume reflects LS Electric’s capabilities in power distribution technology and its ability to meet the demanding requirements of large-scale data center projects.

The project is particularly significant because it involves the integrated supply of switchgear and distribution transformers. As AI data centers consume increasing amounts of electricity, on-site generation systems using fuel cells are gaining attention as a way to support reliable power supply. This trend is also creating opportunities for suppliers of essential power distribution equipment.

LS Electric is accelerating its North American expansion by leveraging its technology, product quality and ability to meet tight delivery schedules. The company is also developing a localized value chain spanning sales, engineering, manufacturing and services through its LS Electric Utah facility and Bastrop Campus in Texas.

Building on the latest contract, LS Electric plans to deepen its strategic partnership with Bloom Energy and pursue additional opportunities in North America, particularly in power infrastructure for AI data centers and fuel cell-based generation facilities.

“ This contract demonstrates our technological capabilities and project execution expertise in supplying integrated power distribution solutions for large-scale data centers,” an LS Electric official said. “We will further strengthen our partnership with Bloom Energy and leverage our local manufacturing and service infrastructure to expand our presence in North America’s rapidly growing AI data center power infrastructure market.”

The South Korean biopharmaceutical company announced on October 13 that the research had been published in the Journal of Immunological Methods, an international scientific journal indexed in the Science Citation Index Expanded (SCIE).

Hunter syndrome is a rare genetic disorder caused by a deficiency of enzymes responsible for breaking down glycosaminoglycans (GAGs) in the body. Enzyme replacement therapy (ERT), which supplies the missing enzyme, is a key treatment approach for the condition.

However, protein-based therapies can trigger immune responses in some patients, leading to the production of anti-drug antibodies (ADAs). Among these, IgE antibodies are of particular concern because they can be associated with acute allergic reactions, including breathing difficulties and a drop in blood pressure.

Accurately identifying these antibodies is therefore important for assessing potential treatment-related hypersensitivity and improving patient safety.

Conventional antibody tests generally determine whether antibodies against a therapeutic drug are present but may not distinguish specific antibody classes. Detecting IgE antibodies is particularly challenging because they circulate in the blood at very low concentrations, making it difficult to establish whether an allergic reaction is directly associated with a particular medication.

GC Biopharma's newly developed assay uses a high-sensitivity enzyme-linked immunosorbent assay (ELISA) to selectively identify IgE antibodies that specifically react with Hunterase, including those present at low concentrations in patients' blood samples.

The company expects the technology to enable more precise assessments of the relationship between Hunterase administration and potential allergic reactions. It could also support more effective monitoring of immune responses during treatment and contribute to broader patient safety management.

Beyond Hunterase, GC Biopharma sees potential for applying the analytical method to other protein-based therapeutics to detect drug-specific IgE antibodies and evaluate immunogenicity, or the ability of a drug to trigger an immune response.

Hong Jung-woon, head of the R&D Quality Management Unit at GC Biopharma, said the assay would provide a foundation for more precise evaluation of drug-specific IgE-mediated hypersensitivity in patients receiving treatment for Hunter syndrome.

He added that the company would continue strengthening its expertise in immunogenicity assessment and analytical technologies to help improve pharmaceutical safety and patient care.

The group announced on October 13 that Shinhan Indonesia Bank had transitioned into a business holding company, officially launching Shinhan Financial Group Indonesia. The move marks the first establishment of a local financial holding company in Indonesia by a South Korean financial group.

The restructuring began in December 2024 following Indonesian financial regulators’ requirement for financial holding company structures. After receiving regulatory approval in September 2025, Shinhan completed the necessary share transfers and related procedures on September 16, with support from the Korean Embassy in Indonesia and the Korea-ASEAN Financial Cooperation Center.

As part of the reorganization, Shinhan Indonesia Bank acquired additional stakes in Shinhan Indo Finance and Shinhan Investment Securities Indonesia, previously held by Shinhan Card and Shinhan Securities, respectively. The bank now holds 50% plus one share in each company, positioning it as the central entity overseeing the group's local financial businesses.

Over the past decade, Shinhan has expanded its Indonesian operations into a comprehensive financial group employing approximately 900 people. Through the new holding company structure, the group plans to strengthen collaboration among its subsidiaries, streamline decision-making and improve operational efficiency.

An inauguration ceremony was held in Jakarta on October 12, attended by Shinhan Financial Group Chairman Jin Ok-dong and key representatives from Indonesian financial authorities. Participants celebrated the launch and discussed the vision for the newly established entity.

On the same day, Shinhan Financial Group signed a memorandum of understanding with Sinar Mas Group, Indonesia's fourth-largest business conglomerate, to explore new growth opportunities in the local market.

The partnership will focus on corporate finance and supply chain finance, combining Sinar Mas's extensive industrial network with Shinhan's financial expertise. The two groups aim to identify new business opportunities and deliver differentiated financial services to local customers.

The agreement represents part of Shinhan's broader strategy to diversify its Indonesian operations following the launch of the holding company. By working with a major local business group, Shinhan expects to expand its corporate client base and develop additional revenue streams.

Jin said the new holding company provides a foundation for managing Shinhan's Indonesian businesses under an integrated framework. He added that cooperation with Sinar Mas would help the group deliver greater value to local customers while contributing to economic development in both South Korea and Indonesia.

The restructuring is expected to strengthen coordination across Shinhan's Indonesian subsidiaries and improve its ability to respond to local market demand. As competition and demand for specialized financial services evolve across Southeast Asia, Shinhan plans to leverage its integrated operating structure and strategic partnerships to pursue further growth in Indonesia.

The South Korean beauty retailer announced on October 12 that its curated retail concept, ‘OLIVE YOUNG K-Beauty Edit,’ had launched at selected Sephora stores and online platforms in Malaysia, Singapore, Thailand and Hong Kong.

The latest rollout covers 48 physical stores, including 18 in Malaysia, nine in Singapore, 13 in Thailand and eight in Hong Kong. Each market features a selection of 12 to 17 Korean beauty brands, with up to approximately 120 products available.

The expansion follows a strategic partnership signed between Olive Young and Sephora in January. Under the collaboration, Olive Young leverages its expertise in identifying promising Korean beauty brands and curating products, while Sephora provides access to its established international retail network.

The strategy is designed to connect emerging K-beauty brands with overseas consumers more efficiently, complementing Olive Young's own retail operations and global online platform.

The partnership has already shown encouraging results in the United States, where the K-Beauty Edit concept was introduced at Sephora in August. By the end of September, average daily sales of participating brands, including Rejuran, Arencia and Torriden, had doubled compared with August.

Building on this momentum, Olive Young is now expanding its reach across Asian markets, where demand for Korean skincare and beauty products continues to create opportunities for international growth.

For its Southeast Asian expansion, Olive Young has tailored its product assortment to regional conditions, including hot and humid weather and strong ultraviolet exposure.

The lineup emphasizes cleansing products, serums and creams from brands such as Dr. Althea, Manyo, Bioheal BOH and Beplain. The selection reflects skincare needs associated with hydration, skin-barrier care and daily protection in tropical climates.

To mark the launch, Olive Young is also hosting a pop-up event at ION Orchard, a premium shopping mall in Singapore, through October 14.

Visitors can receive skin analysis services and beauty samples while exploring featured brands organized around four skincare themes: slow aging, skin-barrier care, pore and blemish management, and hydration.

Through this localized approach, Olive Young aims to offer more than a general selection of Korean cosmetics, presenting products based on consumer preferences and market-specific skincare needs.

Olive Young plans to continue expanding the K-Beauty Edit concept through Sephora's international retail network. In 2027, the company intends to introduce the concept to additional Sephora stores in Canada, Australia, the United Kingdom and the Middle East.

The expansion is expected to create additional distribution opportunities for Korean beauty companies seeking to enter major overseas retail channels. It also allows Olive Young to apply its product-curation expertise and consumer trend analysis to markets beyond its direct retail footprint.

An Olive Young representative said the company was working to broaden consumer access to K-beauty both through its own presence in the United States and through its partnership with Sephora.

The representative added that Olive Young would continue using market-specific consumer data and trend insights to curate products and expand the global customer base for Korean beauty brands.

As K-beauty gains wider international exposure, the company's dual strategy of direct market entry and partnerships with established global retailers is emerging as a key part of its efforts to strengthen the worldwide distribution of Korean beauty products.

The bank announced on October 13 that it had won the Best Private Bank for Innovation award at the 36th Global Wealth Awards 2026, organized by Private Banker International (PBI), a global financial publication specializing in private banking and wealth management. The awards ceremony was held in Singapore on October 8.

Hana Bank also achieved notable success in the individual categories. It received the Regional Private Banker of the Year award for the seventh consecutive year, recognizing excellence among private bankers operating across Asia. In addition, the bank's private banking professionals have won the Rising Star in Private Banking award five times in total.

PBI recognized Hana Bank for its customer-focused wealth management solutions, comprehensive services for ultra-high-net-worth individuals, and innovative financial products and personalized services tailored to clients' needs.

The recognition highlights Hana Bank's continued efforts to strengthen its wealth management capabilities by combining specialized financial expertise with personalized customer service.

<>Comprehensive Wealth Management Drives Recognition

Hana Bank's latest achievement reflects its strategy of providing integrated wealth management services that address the increasingly diverse needs of high-net-worth clients.

The bank has developed specialized solutions tailored to individual financial circumstances and investment objectives. In addition to traditional financial services, it offers a broader range of wealth management support designed to address clients' needs across different stages of life.

As demand for sophisticated wealth management services grows, financial institutions are increasingly focusing on personalized investment strategies and integrated solutions for affluent customers. Hana Bank has sought to differentiate itself by combining financial services with non-financial advisory capabilities.

The latest award recognizes the bank's approach to developing customized solutions while adapting to changes in financial markets and evolving customer expectations.

Hana Bank has continued to build its international reputation in private banking over the past decade. Since becoming the first commercial bank in South Korea to receive the Best Private Bank in East Asia award in 2012, the bank has accumulated a total of 15 wins at the Global Wealth Awards.

The latest recognition underscores both the bank's institutional capabilities and the expertise of its private banking professionals.

The consecutive awards in the Regional Private Banker of the Year category also highlight the consistency of Hana Bank's talent development and client service capabilities. Meanwhile, its repeated success in the Rising Star in Private Banking category reflects its efforts to cultivate the next generation of wealth management specialists.

<>Strengthening Its Position in the Global Wealth Management Market

Founded in 1987, the Global Wealth Awards is one of the longest-running award programs in the global private banking industry. Organized annually by PBI, the awards recognize financial institutions and professionals for excellence in wealth management and private banking services.

Hana Bank plans to continue expanding its integrated wealth management capabilities, with a focus on personalized services that reflect clients' financial goals, asset levels, and life stages.

The bank also aims to strengthen its offerings beyond conventional financial products by incorporating a wider range of non-financial services into its wealth management solutions. Through this approach, it seeks to build long-term relationships with customers and reinforce its role as a trusted financial partner throughout their lives.

A Hana Bank Wealth Management Group representative said the bank was proactively responding to rapidly changing market conditions through customer-oriented wealth management services.

The representative added that Hana Bank would continue enhancing its comprehensive capabilities across financial and non-financial services to fulfill its role as a lifelong financial partner for customers.

LS Materials announced on October 13 that it had signed a contract worth approximately $51.5 million, or around KRW 69 billion, to supply UCs for AI data center applications. The deal represents the largest export contract ever secured by a South Korean company in the ultracapacitor industry.

The contract value is approximately 2.4 times LS Materials' separate revenue of KRW 29 billion last year, highlighting the potential scale of the new business opportunity.

Deliveries are scheduled to begin in April 2027 and continue over approximately one year. The company is also discussing additional supply agreements with the customer, raising the prospect of further order growth.

The latest deal follows an initial supply agreement worth approximately $4.7 million, or KRW 6.3 billion, signed with the same company in July. Shipments under that agreement are scheduled to begin in November this year.

Including the latest contract, LS Materials' cumulative orders from the customer have reached approximately $56 million. The company expects total orders to exceed $100 million if additional agreements are secured by the end of the year.

The contract marks LS Materials' first supply agreement for ultracapacitors used in AI data center infrastructure. The company has traditionally focused on industrial applications, including wind turbines, automated guided vehicles (AGVs), and uninterruptible power supply (UPS) systems.

Entering the AIDC market will allow LS Materials to diversify its customer base and pursue new growth opportunities in the rapidly expanding AI infrastructure sector.

Ultracapacitors are energy storage devices capable of rapidly charging and discharging large amounts of electricity. They help stabilize power output by responding to sudden fluctuations in electricity demand, making them particularly useful in power systems that require reliable and consistent energy delivery.

In hydrogen fuel cell systems, UCs can compensate for rapid changes in power demand, helping stabilize output and improve overall system performance.

The rapid growth of generative AI and high-performance computing is driving increased electricity demand at data centers worldwide. As operators seek reliable power infrastructure to support increasingly demanding workloads, interest in alternative power generation technologies, including hydrogen fuel cells, is also growing.

This trend is expected to create additional demand for ultracapacitors capable of managing short-term power fluctuations in fuel cell-based generation systems.

The market for medium- and large-sized UCs requires advanced technological capabilities, including high output performance, reliability, and system integration expertise. With only a limited number of global suppliers possessing the necessary capabilities, LS Materials sees opportunities to strengthen its position as demand for AI data center power solutions expands.

The company plans to leverage its expertise in medium- and large-sized ultracapacitors to pursue further business opportunities in the global market.

The latest contract represents more than a significant export achievement. It also signals a strategic shift toward higher-growth applications linked to AI infrastructure and next-generation power systems.

By applying its existing UC technology to hydrogen fuel cell systems for AI data centers, LS Materials aims to establish a new revenue stream while strengthening relationships with global customers.

The company will focus on fulfilling the contracted deliveries on schedule while continuing negotiations for additional orders. Successful execution and further customer agreements will be important factors in determining the pace of its expansion into the AIDC market.

Hong Young-ho, CEO of LS Materials, said the agreement was significant because it established a new growth foundation through the company's entry into the hydrogen fuel cell market for AI data centers.

He added that LS Materials would strengthen cooperation with customers and expand its global presence by leveraging its technological capabilities in medium- and large-sized ultracapacitors.

LS Materials plans to build on the latest agreement to expand its role in AI data center power infrastructure, positioning its UC technology to capture emerging demand for stable and reliable electricity supply systems.

Microsoft expects Maia 300 to deliver approximately 40% higher AI throughput per megawatt than leading GPU-based systems. The measurement is based on tokens, the basic units that AI models process and generate. A server rack containing 96 accelerators is projected to deliver up to seven times the throughput of its predecessor, Maia 200. These figures represent Microsoft's performance targets and have yet to be independently verified under real-world operating conditions.

AI Chip Competition Expands Beyond Computing Power

Scott Guthrie, Microsoft's executive vice president of Cloud and AI, emphasized in the company's announcement that AI is becoming a foundational layer for applications and business processes. As enterprises move beyond conversational AI toward agents capable of performing complex tasks, data centers face growing computational workloads and electricity demand. Microsoft said Maia 300 is designed to support advanced reasoning, long-context processing and multistep AI operations.

As AI services expand, operating costs are becoming as important as semiconductor acquisition costs. Electricity, cooling, memory and networking all contribute to the economics of large-scale AI infrastructure. Processing more requests with the same amount of power can reduce the cost per service and improve data center profitability. Microsoft has therefore designed Maia 300 as an integrated system combining processors, memory and networking, using TSMC's N2P manufacturing process and HBM4 high-bandwidth memory to improve computational and data-processing efficiency.

Reducing Nvidia Dependence Creates New Opportunities for Korean Chipmakers

Microsoft's expansion of proprietary AI chips could reshape a supply chain dominated by Nvidia. Major cloud providers continue to purchase Nvidia GPUs while developing their own accelerators to diversify supply and control costs. AI inference workloads, in particular, vary in their computing and memory requirements, creating opportunities for specialized chips. Rather than replacing Nvidia products entirely, the strategy allows cloud providers to select hardware suited to specific services.

The shift could also create opportunities for South Korea's semiconductor industry. Maia 300's adoption of HBM4 highlights the growing importance of high-bandwidth memory not only in AI training but also in inference infrastructure. For Samsung Electronics and SK hynix, developing memory products and advanced packaging technologies tailored to proprietary AI accelerators could become an increasingly important competitive priority. However, actual suppliers, shipment volumes and pricing arrangements remain to be confirmed.

Greater power efficiency does not necessarily mean lower total electricity consumption. Even as individual chips become more efficient, rapidly growing demand for AI services could increase overall data center power usage. Evaluating AI infrastructure economics therefore requires more than comparing semiconductor performance. Electricity procurement costs, equipment utilization and the time needed to recover capital investment must also be considered.

The commercial performance of Maia 300 will become clearer after deployment begins in 2027. The key questions are whether Microsoft's projected efficiency gains translate into lower operating costs and whether its proprietary accelerators prove more economical than externally supplied GPUs. As AI infrastructure expands, competitive advantage will increasingly depend not only on access to computing hardware but also on how efficiently companies convert electricity and capital investment into reliable, revenue-generating AI services.

Read the Korean version

Aston Science, a clinical-stage cancer vaccine developer, announced on October 12 that the U.S. Food and Drug Administration (FDA) had granted Fast Track designation to AST-301, its investigational HER2-targeted cancer therapeutic vaccine, for the treatment of HER2-expressing gastric cancer. The designation comes approximately four months after AST-301 received orphan drug designation in June.

Fast Track is designed to facilitate the development of medicines for serious conditions with unmet medical needs. It supports more frequent meetings and written communication with the FDA during development. It also provides an opportunity for rolling review, allowing completed sections of a marketing application to be submitted for review as they become available. A Fast Track product may also qualify for priority review or accelerated approval if the relevant criteria are met, potentially shortening the path to commercialization.

AST-301 is a plasmid DNA-based cancer therapeutic vaccine that uses the intracellular domain (ICD) of HER2 as its antigen. The investigational immunotherapy is designed to induce HER2-specific T-cell responses to eliminate cancer cells. Aston Science has completed patient enrollment and AST-301 dosing in its Phase 2 trial involving patients with HER2-expressing gastric cancer and is continuing to monitor outcomes. The company expects to announce results for key endpoints in October.

According to Aston Science, HER2-expressing gastric cancer accounts for more than 35% of all gastric cancer cases and is associated with a relatively poor prognosis. The company said this underscores demand for new treatments using immunological approaches such as AST-301.

“The successive orphan drug and Fast Track designations for AST-301’s gastric cancer indication demonstrate the global competitiveness of a Korean company in the regulatory development of cancer therapeutic vaccines,” said Eunyoung Choi, head of regulatory affairs at Aston Science. “They also recognize AST-301’s development potential as a treatment that could address unmet medical needs in HER2-expressing gastric cancer.”

Following the Fast Track designation, Aston Science plans to accelerate AST-301’s subsequent clinical development program through closer and more timely engagement with the FDA. The company will also consider pursuing an Early Access Program (EAP) in the United States. “Our goal is to advance toward the submission of a Biologics License Application (BLA),” a company representative said.

Read the Korean version

The company is participating in FAKUMA 2026, an international plastics processing trade fair held in Friedrichshafen, Germany, from October 12 to 16.

Under the theme of automotive, robotics and medical applications, Hyosung Chemical is showcasing POKETONE-based automotive connectors, gears, drive components and medical parts.

POKETONE offers impact and wear resistance, chemical resistance and low moisture absorption, helping components maintain dimensional stability. Its properties, including strength, wear resistance and flame retardancy, can also be tailored to specific applications.

Hyosung Chemical is exploring the material’s use in humanoid robot joint gears and shaft-support components, which must withstand repeated movement and friction. The company aims to build on existing applications in service robots and identify additional demand in Europe’s automotive and industrial manufacturing sectors.

On October 14, Hyosung Chemical will hold a technical seminar titled “Proven Worldwide, Engineered Forward” to discuss the material’s capabilities and potential applications.

CEO Lee Cheon-seok said the company would expand POKETONE’s applications across electric vehicles, robotics and medical industries through customer collaboration and tailored technical support, responding to growing demand for high-performance and sustainable materials.

The South Korean heating and energy solutions company plans to use the facilities to develop products tailored to local housing conditions, improve technical support and quality management, and provide specialized training for distributors, installers and other industry partners.

The openings mark a new phase in Kyungdong Navien’s North American strategy as the company seeks to build on its established water heater and boiler businesses while expanding into a broader range of climate control and energy-efficient home solutions.

Ceremonies held in New Jersey on Oct. 6 and Irvine on Oct. 9 local time attracted approximately 170 industry representatives, including major North American distributor Johnstone Supply and installation and contracting company Indoor Weather, highlighting interest from local industry partners.

The Innovation Centers will serve as technical hubs for product development, testing and quality improvement. The facilities will evaluate product performance through laboratory testing and field trials, analyze issues identified in actual installations, and incorporate the findings into product and component improvements.

Kyungdong Navien also plans to strengthen collaboration with academic institutions and research organizations while maintaining close coordination with its research and development teams in South Korea. The centers will support the development of products suited to North American market requirements and help facilitate certification processes in the United States.

The Training Academies will focus on improving the technical capabilities of installers, contractors, distributors and sales representatives. Dedicated classrooms and practical training areas will allow participants to receive instruction ranging from product fundamentals and demonstrations to installation procedures, system inspections and troubleshooting.

Training programs will initially build on the company’s existing water heaters, boilers and combination systems before expanding to HVAC equipment and water treatment solutions. The broader curriculum is intended to prepare local partners to install, maintain and support a growing range of products.

Kyungdong Navien plans to operate the Innovation Centers and Training Academies as an integrated system. Feedback gathered during training and from field installations will be shared with product development teams to identify opportunities for improvements. Updated products and technologies will then be introduced to the field, creating a feedback loop between development, installation and customer support.

The company expects this approach to improve product reliability, accelerate responses to local market needs and strengthen relationships with distribution and installation partners.

The new facilities also reflect Kyungdong Navien’s ambition to expand beyond its established position in North American water heaters and boilers. The company is seeking to develop a broader portfolio of HVAC and energy-related solutions tailored to regional climate conditions, housing structures and installation requirements.

Products recently introduced to the North American market include heat pumps, heat pump water heaters and water treatment systems. Kyungdong Navien also plans to develop additional solutions, including hybrid systems that combine different heating technologies, as potential new growth drivers.

By using the centers to test products under local conditions and gather feedback from installers, the company aims to improve the performance and suitability of its new offerings before expanding their market reach.

Kyungdong Navien is also carefully evaluating the possibility of establishing a manufacturing base in the United States. Local production could help improve supply stability and shorten response times as the company expands its product portfolio and customer base. However, the plan remains under consideration, with specific investment details and a timeline yet to be finalized.

A Kyungdong Navien official said the new facilities would serve as strategic hubs for translating feedback from North American customers and industry partners into product development and improved customer value.

“Building on the data and technical capabilities accumulated at these centers, we will develop differentiated HVAC products, including hybrid systems, while pursuing the establishment of a U.S. manufacturing base to create a new turning point for our business,” the official said.

The openings represent a broader shift in Kyungdong Navien’s North American operations, from supplying established heating products toward offering a wider range of integrated climate control and water-related solutions.

As competition in the HVAC market intensifies and demand for energy-efficient technologies grows, the company’s ability to localize product development, support installation partners and maintain reliable supply will be critical to its expansion.

The effectiveness of the new facilities will ultimately depend on how successfully Kyungdong Navien converts local technical expertise and customer feedback into competitive products, stronger distribution relationships and sustainable growth in the North American market.

The company said on Oct. 12 that its board of directors had approved a plan to pursue the name change, reflecting its ambition to build on nearly six decades of pharmaceutical expertise while developing new growth opportunities in global markets.

Dongkook Pharmaceutical has primarily focused on the domestic pharmaceutical market. Under its new strategic direction, the company aims to broaden its business scope to encompass prevention, treatment, health and beauty, positioning itself as a science-based health and beauty company with a stronger international orientation.

The proposed name combines the company’s established identity with its ambitions for future growth. “DK” represents the company’s 58-year history, scientific credibility and enduring corporate foundation, while “Centellian” reflects its commitment to creating new, forward-looking value through accumulated expertise.

The company said the change is intended to communicate its broader business vision and strengthen its connection with customers around the world.

By expanding beyond its conventional pharmaceutical operations, Dongkook Pharmaceutical aims to build a more diversified business portfolio that integrates its existing capabilities with opportunities across the health and beauty sectors. The strategy is also expected to support its efforts to establish a stronger brand presence in international markets.

The proposed rebranding marks a strategic shift in how the company seeks to position itself for future growth, while retaining the scientific expertise and trust developed over decades in the pharmaceutical industry.

To proceed with the change, Dongkook Pharmaceutical announced that it will convene an extraordinary general meeting of shareholders on Nov. 27. The meeting will consider amendments to the company’s articles of incorporation required for the name change.

The proposed rebranding remains subject to the relevant corporate approval procedures.

A company official said the new name reflects the company’s determination to preserve its longstanding foundations while creating new value through its accumulated expertise.

The proposed name change underscores Dongkook Pharmaceutical’s ambition to evolve from a domestically focused pharmaceutical company into a more diversified health and beauty business with global reach. Its next challenge will be to translate that vision into tangible growth through business expansion, international market development and stronger brand recognition.

LG Electronics said on Oct. 11 that it held a gala at the Guggenheim Museum in New York on Oct. 8 local time to celebrate the milestone and showcase its expanded LG SIGNATURE lineup. The strategy centers on combining sophisticated design with artificial intelligence (AI)-powered functionality while making the brand accessible to a wider range of premium consumers.

The expanded portfolio includes kitchen appliances such as oven ranges, wall ovens, range hoods and cooktops, bringing the brand’s lineup to 10 product categories. The move strengthens LG’s presence in the kitchen appliance segment and allows the company to offer a more comprehensive selection for consumers seeking a consistent design aesthetic throughout their homes.

<>Three Design Collections Target Diverse Consumer Preferences

LG SIGNATURE has traditionally emphasized minimalist design, premium materials and advanced technology. For its next phase of growth, LG Electronics is dividing the lineup into three distinct design collections: Iconic, Tailored and Seamless.

The Iconic collection combines metallic finishes with gold accents to create a distinctive, upscale appearance. Tailored offers a wider selection of colors, allowing consumers to personalize their appliances according to individual tastes and interior design preferences. Seamless focuses on integration, with appliances designed to sit flush with surrounding cabinetry and minimize protruding elements.

Together, the three collections reflect LG’s effort to move beyond a one-size-fits-all luxury aesthetic. By offering different interpretations of premium design, the company aims to appeal to customers with varying preferences while maintaining a consistent brand identity.

The built-in approach is particularly relevant to LG’s North American growth strategy. Appliances that integrate with cabinetry and complement furniture can help create a unified interior, making them attractive to homeowners seeking a coordinated, high-end living environment.

LG is positioning SIGNATURE not simply as a collection of premium appliances, but as a way to enhance the overall appearance and functionality of residential spaces.

<>AI Features Add Intelligence to the Premium Experience

Alongside design, LG is strengthening the lineup’s performance and AI-based convenience features.

One example is the LG SIGNATURE oven range equipped with Gourmet AI. An internal camera identifies ingredients and recommends suitable cooking methods, helping users select appropriate settings and simplify meal preparation.

The system is designed to make cooking more intuitive by reducing the need for users to determine every setting manually. AI-powered monitoring and recommendations can also help users manage the cooking process more conveniently.

The technology reflects a broader shift in the premium appliance market, where differentiation increasingly depends on the combination of industrial design, connected services and intelligent functionality rather than hardware specifications alone.

LG has been expanding AI capabilities across its broader appliance portfolio, seeking to make everyday household tasks more convenient through automated recognition, personalized recommendations and connected-device features.

For LG SIGNATURE, the challenge is to translate those technologies into a premium experience that complements the brand’s emphasis on understated design and refined living spaces.

<>Retail Strategy Shifts Beyond Flagship Stores

LG Electronics is also changing how it presents and sells the brand in the United States.

Previously, LG SIGNATURE products were displayed primarily in flagship stores to reinforce their premium positioning. Beginning with the newly introduced products, the company plans to expand displays to regular stores operated by major retail partners.

The shift is intended to increase opportunities for consumers to see and experience the products firsthand. While flagship locations remain important for communicating the brand’s identity, broader retail distribution can help LG reach customers who may not regularly visit dedicated premium showrooms.

The strategy also reflects the need to balance exclusivity with accessibility. A broader retail presence can increase consumer awareness and support purchasing decisions, particularly for kitchen appliances that require careful consideration of design, dimensions and installation requirements.

By expanding both its product range and retail touchpoints, LG aims to turn the brand’s established premium image into stronger commercial momentum.

<>North America Remains a Key Growth Market

The United States is an important market for LG’s premium appliance business, particularly as the company seeks to strengthen its position in built-in kitchen products.

Consumers in the premium segment often consider how appliances fit into the overall design of a home, alongside factors such as performance, reliability and convenience. LG’s expanded lineup is intended to address these requirements through coordinated design options and advanced features.

The company is also highlighting the performance of its laundry products. LG SIGNATURE washing machines and dryers were included in Consumer Reports’ list of the 15 best washer-dryer sets of 2026. The LG SIGNATURE washer, model WM9900HSA, received 87 points, the highest score among approximately 100 products evaluated, according to the company.

Such recognition provides an additional point of differentiation as LG seeks to compete in a market where consumers weigh brand reputation, product performance and long-term value when making premium purchases.

“Over the past 10 years, LG SIGNATURE has set a new standard for premium home appliances with restrained designs focused on the essentials of appliances and innovative technology,” said Kwak Do-young, vice president and head of LG Electronics’ North America region.

“Going forward, we will continue to provide North American customers with a differentiated experience through a premium lineup that combines designs that elevate living spaces with AI-powered convenience,” he added.

As LG SIGNATURE enters its second decade in the U.S. market, the company is pursuing a broader strategy that combines design differentiation, intelligent features and wider distribution. The expanded 10-category lineup gives LG a stronger foundation to compete in North America’s premium and built-in segments, while the three design collections offer greater flexibility for consumers seeking appliances that complement their homes.

The key test will be whether LG can convert the brand’s established premium positioning and expanded retail presence into sustained sales growth in an increasingly competitive market.

The South Korean conglomerate is expanding cooperation with Microsoft across AI data centers, AI transformation (AX) and physical AI, while LG Electronics and LG Energy Solution are strengthening their positions in Nvidia’s AI factory ecosystem through cooling and energy storage technologies, respectively.

The strategy centers on combining the capabilities of LG’s affiliates to provide integrated solutions for AI infrastructure, linking cooling systems, power management, energy storage and IT services. As global technology companies increase investment in AI computing capacity, LG aims to turn rising demand for data center infrastructure into a new engine of growth.

<>LG and Microsoft Expand Strategic Cooperation Across Three AI Areas

Recently, Koo met with Microsoft CEO Satya Nadella and other senior executives at Microsoft’s headquarters in Redmond, Washington, on Oct. 21 local time. The two companies agreed to expand their strategic partnership across three areas: AI data center infrastructure, enterprise AI transformation and physical AI.

In AI data centers, the companies will explore business opportunities to supply LG’s cooling, power and IT infrastructure solutions in line with Microsoft’s global data center expansion plans.

The initiative reflects the growing importance of infrastructure beyond computing chips and servers. As AI workloads become more intensive, data centers require greater power capacity and more advanced thermal management systems to maintain operational stability and improve energy efficiency.

LG believes its broad portfolio of businesses will allow it to respond to these requirements by bringing together technologies and solutions from multiple affiliates rather than approaching each infrastructure component as a standalone product.

The partnership will also cover enterprise AI transformation. LG plans to introduce Microsoft’s AI and cloud platforms, including Copilot, to office workers across the group. The initiative is intended to improve productivity by applying AI to routine tasks such as document preparation, information retrieval and data analysis.

In physical AI, the companies will explore the use of Microsoft’s high-performance cloud infrastructure to develop and train AI models for real-world industrial applications, including manufacturing, logistics and mobility.

Physical AI combines AI models with robots and other machines that perceive and interact with the physical world. Its development requires substantial computing resources, extensive training data and infrastructure capable of supporting complex simulations and model training.

By combining LG’s industrial and manufacturing capabilities with Microsoft’s cloud infrastructure, the companies aim to identify opportunities to accelerate the development and deployment of AI in industrial environments.

The latest agreement builds on earlier cooperation. In December last year, five LG affiliates — LG Electronics, LG Innotek, LG Energy Solution, LG Uplus and LG CNS — visited Microsoft’s headquarters to showcase technologies related to AI data centers at an event known as the AIDC Tech Show. The companies continued working-level discussions over the following months before expanding their cooperation at the group level.

<>LG Electronics Secures Nvidia Approval for 2.5-Megawatt Cooling System

LG is also deepening its relationship with Nvidia, a central player in the global AI computing ecosystem.

Koo met Nvidia CEO Jensen Huang in June and held a meeting with senior executives. In August, he signed a memorandum of understanding at Nvidia’s headquarters in Santa Clara, California, to pursue strategic business cooperation in robotics, AI factories and mobility.

The collaboration is increasingly reflected in the activities of LG’s individual affiliates.

LG Electronics announced that Nvidia had approved its 2.5-megawatt cooling distribution unit (CDU) for use in AI factory infrastructure. The approval falls under Nvidia’s “DSX Ready” program, which recognizes products that meet requirements for its AI factory ecosystem.

A CDU distributes and manages cooling liquid used to remove heat generated by high-performance computing equipment. As AI servers become more powerful and rack densities increase, liquid cooling is gaining importance as a means of managing heat more efficiently than conventional air-cooling systems alone.

LG Electronics previously received Nvidia approvals for its 600-kilowatt CDU in July and its 1-megawatt model in September. Approval of the 2.5-megawatt unit expands its portfolio of products designed for high-capacity AI infrastructure.

The development also represents an extension of LG Electronics’ established heating, ventilation and air-conditioning expertise into the rapidly growing AI data center market.

As AI computing demand rises, cooling performance is becoming a key consideration in data center design. Efficient thermal management can help operators maintain system reliability while controlling energy consumption and operating costs.

The Nvidia approval provides LG Electronics with a stronger basis for competing in this market. However, product qualification does not itself guarantee commercial orders or revenue. The scale of future business will depend on adoption by data center operators, project requirements and subsequent supply agreements.

<>LG Energy Solution Joins Nvidia’s AI Factory Ecosystem With Energy Storage Technology

LG Energy Solution is expanding the group’s presence in AI infrastructure through battery-based energy storage.

The company has been selected as a partner for Nvidia’s DSX Ready battery energy storage system (BESS) program, becoming the only South Korean battery manufacturer selected for the ecosystem, according to LG.

The approved product is the JF2 AC Link, a high-capacity lithium iron phosphate (LFP)-based energy storage solution. LFP batteries are widely used in stationary storage applications because of their durability, long cycle life and cost competitiveness.

Energy storage systems can play an important role in data center power management by storing electricity and supplying it when needed. As AI facilities require increasingly large and stable power supplies, BESS technology offers operators an additional tool for managing fluctuations in electricity demand and improving operational flexibility.

The partnership could create new opportunities for LG Energy Solution beyond its traditional electric vehicle battery business. Growing investment in AI infrastructure is expected to increase demand not only for computing equipment but also for supporting energy systems capable of delivering reliable power at scale.

For LG Energy Solution, Nvidia’s qualification provides an entry point into a strategically important ecosystem. Actual commercial results, however, will depend on the adoption of its products in individual projects, customer investment decisions and the terms of future supply contracts.

Together, LG Electronics’ cooling technology and LG Energy Solution’s storage capabilities illustrate how the group is positioning multiple affiliates to address different infrastructure requirements within the same AI data center market.

<>‘One LG’ Seeks to Turn Individual Technologies Into Integrated Solutions

Industry observers expect the expansion of global AI infrastructure investment to increase demand for suppliers capable of coordinating multiple systems, rather than providing isolated components.

An AI data center requires a combination of high-performance computing equipment, power infrastructure, cooling systems, networking and IT services. These elements must operate together to support demanding workloads while maintaining efficiency and reliability.

LG’s portfolio gives it the potential to address several of these requirements through a coordinated approach.

LG Electronics brings cooling and HVAC expertise, LG Energy Solution contributes battery and energy storage technology, and LG CNS provides IT infrastructure and systems integration capabilities. Other affiliates, including LG Innotek and LG Uplus, may also contribute through their respective technologies and businesses.

By coordinating these capabilities, LG aims to strengthen its position as a broader AI infrastructure solutions provider rather than competing solely as a supplier of individual products.

The approach could also help the group engage more effectively with global technology companies whose AI infrastructure projects require multiple categories of equipment and services.

Microsoft and Nvidia offer different but complementary opportunities. Microsoft’s partnership with LG spans cloud infrastructure, enterprise AI adoption and industrial AI development. Nvidia’s ecosystem provides a route for LG’s cooling and energy storage products to be considered for AI factory deployments.

The combination gives LG opportunities to participate across several layers of the AI value chain, from the infrastructure that supports computing to the systems that manage its power and heat requirements.

Nevertheless, the commercial impact will depend on LG’s ability to convert technical qualifications and strategic partnerships into project wins, supply agreements and recurring revenue.

<>Koo Takes a Direct Role in Building Global AI Partnerships

Koo has increasingly engaged directly with executives at major global AI companies, seeking to establish group-wide relationships that can create opportunities for LG’s affiliates.

The latest agreement with Microsoft reflects this approach, bringing together infrastructure, enterprise software adoption and industrial AI under a broader strategic framework.

For LG, the challenge is no longer limited to developing competitive individual products. The group must also integrate its technologies, secure positions within global technology ecosystems and deliver solutions that address customers’ infrastructure requirements as a whole.

The “One LG” strategy is designed to connect these capabilities across corporate boundaries and strengthen the group’s ability to respond to emerging market demand.

As investment in AI computing infrastructure continues to reshape the technology industry, cooling efficiency, power reliability and energy management are becoming increasingly important alongside semiconductor performance.

LG’s expanding partnerships with Microsoft and Nvidia offer a potential route to capitalize on these trends. Whether they translate into a material new growth engine will depend on the pace of commercial adoption and the group’s ability to turn collaboration into measurable business results.

For Koo, the priority is to ensure that LG’s diverse businesses operate as a coordinated force in the global AI market, transforming the group’s broad industrial capabilities into a competitive advantage in the next phase of AI infrastructure development.

Held at the Chrysalis Hall of Toronto Metropolitan University at 3 a.m. KST on October 11, the Americas Cup brought together eight players who advanced through regional qualifiers. Competing in a single-elimination bracket with best-of-five matches, the players battled for three coveted spots at the World Finals.

Despite entering the competition ranked third in pre-tournament predictions, SAPPYYY delivered a series of dominant performances to claim the title. In the quarterfinals, SAPPYYY swept TRUEWHALE 3–0, overcoming the top-ranked contender who had won the Americas regional championship in both 2023 and 2024. SAPPYYY continued the impressive run in the semifinals, defeating ZEDORUS by the same score to advance to the final.

The championship match proved to be a much closer contest. Facing ZEZAS, SAPPYYY battled through five sets before securing a 3–2 victory and lifting the trophy. The result was particularly memorable for local fans, as a Toronto-born competitor became the Americas champion on home soil.

ZEZAS finished as runner-up, while MONAGUA secured third place by defeating ZEDORUS 3–0 in the third-place playoff. The three players advancing to the World Finals from the Americas are SAPPYYY, ZEZAS, and MONAGUA.

The top four finishers—SAPPYYY, ZEZAS, MONAGUA, and ZEDORUS—received medals and a share of the tournament's total prize pool of $20,000. The champion took home $10,000.

With the Americas Cup concluded, nine of the 12 competitors for the SWC2026 World Finals have now been determined. The remaining three places will be awarded to the top three finishers at the Asia-Pacific Cup, scheduled to take place in Bangkok, Thailand, on October 17.

The 12 qualified players will meet at KBS Arena in Seoul on November 14 to compete for the title of SWC's 10th world champion.

Launched in 2017, SWC is Com2uS's premier global esports competition based on its hit mobile game Summoners War. The annual tournament has grown into a major international event, bringing together players and fans from around the world. This year's edition marks the competition's 10th anniversary.

Fans who missed the Americas Cup can watch the matches on the official Summoners War esports YouTube channel. Further tournament updates and details are available through the official Summoners War community forum.

The South Korean pharmaceutical company announced on October 12 that Darbecron, developed under the project name DA-3880, is now available in the UAE. Local regulatory approval and commercialization are being handled by Julphar, a pharmaceutical company based in the UAE.

Dong-A ST signed a licensing agreement with Julphar in November 2024, granting the company rights to develop and commercialize DA-3880 across 18 MENA countries. Julphar secured marketing authorization for Darbecron in the UAE in June this year and has since completed its local launch.

The UAE debut represents an important step in Dong-A ST’s efforts to expand its international biosimilar business. The two companies plan to pursue additional regulatory approvals and gradually extend sales to other major markets across the MENA region.

Darbecron is a biosimilar based on darbepoetin alfa, a second-generation erythropoiesis-stimulating agent used to treat anemia in patients with chronic kidney disease and anemia caused by chemotherapy. Dong-A ST is pursuing market-specific development and commercialization strategies to broaden the product’s global reach.

The company has already established a commercial presence in Japan. In 2014, Dong-A ST signed a licensing agreement with Japan’s Sanwa Kagaku Kenkyusho (SKK) for the development and marketing of DA-3880 in the Japanese market. The product received manufacturing and marketing approval in 2019 and is sold under the name “Darbepoetin Alfa BS Injection Sanwa.”

Sales in Japan have grown steadily, reaching KRW 16.5 billion in 2024 and KRW 26.7 billion in 2025, demonstrating the product’s established market presence.

Dong-A ST is developing and marketing DA-3880 as a biosimilar to Nesp in Japan, the MENA region and Türkiye, in line with local market conditions and regulatory strategies. In major markets including the United States and Europe, the company is developing the product as a biosimilar to Aranesp.

Dong-A ST said it would continue working with Julphar to secure additional approvals and expand commercialization across the MENA region. The company also plans to advance development activities aimed at entering major global markets, including the United States and Europe.

The UAE launch builds on the product’s commercial track record in Japan and provides a foundation for further international expansion. Through regional partnerships and market-specific strategies, Dong-A ST aims to strengthen its position in the global biosimilar market.

The investigation centers on the transfer of treasury shares to an in-house employee welfare fund, which enabled the shares to carry voting rights. Investigators are examining whether the decision was a legitimate employee welfare measure or an attempt to strengthen the controlling shareholder’s position.

According to reports, the Financial Crime Investigation Unit of the Seoul Metropolitan Police Agency’s Metropolitan Investigation Division recently conducted a search and seizure at Hanjin KAL. Cho and Hanjin KAL CEO Ryu Kyung-pyo were reportedly identified as suspects in connection with alleged breaches of fiduciary duty under South Korea’s Act on the Aggravated Punishment of Specific Economic Crimes. The allegations remain under investigation, and no finding of guilt has been established.

Police are reportedly examining a 1999 Supreme Court ruling involving former Kia Group Chairman Kim Sun-hong as a potential precedent. In that case, Kia Group executives provided approximately KRW 32.2 billion in company funds to an employee organization known as the Management Development Committee, which used the money to purchase company shares amid a threat to the group’s management control.

Prosecutors argued that the funding was intended primarily to defend management control rather than to promote employee welfare. The courts found the executives criminally liable for breach of fiduciary duty, and Kim received a four-year prison sentence that was upheld by the Supreme Court.

Investigators are examining whether the Kia case shares key legal and factual similarities with Hanjin KAL’s transaction. Both cases involved company assets being used in connection with employee organizations and the acquisition or activation of voting rights. However, there is an important distinction: Kia provided cash for share purchases, while Hanjin KAL transferred shares it already held.

On May 15, 2025, Hanjin KAL transferred 440,044 treasury shares to its employee welfare fund. Based on the share price at the time of the board decision, the shares were valued at approximately KRW 66.3 billion, representing around 0.66% of the company’s total outstanding shares.

Treasury shares generally do not carry voting rights while held by the issuing company. When transferred to an employee welfare fund, however, those shares may acquire voting rights, potentially affecting shareholder decisions and the balance of power in corporate governance.

The size of the contribution has also attracted scrutiny. According to reports, Hanjin KAL had 25 full-time employees. Dividing the estimated value of the transferred shares by that headcount would amount to more than KRW 2.6 billion per employee. This is a simple comparison rather than an estimate of individual benefits, since welfare fund assets are not necessarily distributed equally among employees. Nevertheless, investigators are expected to examine whether the contribution was proportionate to the company’s employee welfare needs.

Another central question is why the company chose to contribute treasury shares rather than cash or other assets. Police are examining whether executives understood that the transfer could activate voting rights, whether management control was discussed during the decision-making process, and whether the transaction had a reasonable business and employee welfare rationale.

The investigation is expected to focus on the purpose of the transfer, the potential impact on the company and the involvement of senior executives. Police are reviewing board meeting minutes, internal reports and other materials obtained during the search to establish who proposed the transaction, when Cho was informed and how the decision was approved.

Investigators are also expected to examine how the welfare fund exercised the voting rights attached to the shares and who influenced those decisions. Interviews with Hanjin KAL employees and key board participants may follow as the review of seized materials progresses.

The central legal issue is whether the transfer represented a legitimate corporate decision or an improper use of company assets that benefited a particular individual’s management control. Establishing a breach of fiduciary duty would require evidence supporting the relevant legal elements, including the executives’ conduct, intent and any resulting harm to the company.

The case could become a significant test of how South Korean authorities assess the use of treasury shares and employee welfare funds in disputes involving corporate control. Whether the Kia Group precedent can be applied to Hanjin KAL will depend on the specific circumstances of the transaction and the evidence gathered regarding the executives’ intentions and decision-making.

Police are continuing to analyze the materials to determine whether the transfer was intended to serve employee welfare or to influence the company’s governance structure. The extent of Cho’s involvement and the circumstances surrounding the board’s decision are likely to be central to the next stage of the investigation.

Samsung SDS is expanding its AI business across data centers, cloud services and manufacturing systems. The company recently agreed to invest in U.S. AI infrastructure company Helix alongside other Samsung affiliates and is developing a platform that connects production equipment with robots. Its cloud business recorded double-digit revenue growth in the second quarter. These initiatives share a common direction: supplying the computing resources AI requires and applying the technology to enterprise systems and manufacturing operations.

At REAL Summit 2026 in Seoul last month, Samsung SDS outlined a strategy combining AI infrastructure, platforms and solutions with consulting, development, implementation and operations. It aims to support customers from AI adoption through ongoing operations, taking account of their workflows and industry requirements. In manufacturing, the company is pursuing Robotics Transformation (RX), connecting robots with existing production systems. The plan extends its AI business from computing infrastructure and enterprise software to physical production environments.

Cloud Growth and Global AI Infrastructure Investment

According to preliminary second-quarter results announced on July 30, Samsung SDS recorded revenue of KRW 3.7178 trillion and operating profit of KRW 231.8 billion. These figures rose 5.9% and 0.7%, respectively, from a year earlier. Cloud revenue increased 17% to KRW 779.4 billion, while total IT services revenue grew 5% to KRW 1.7625 trillion. The results show cloud revenue growing faster than the broader IT services business.

Within cloud services, both infrastructure provision and customer system operations expanded. Revenue from the cloud service provider (CSP) business, which supplies the company’s own cloud infrastructure, rose 24% year on year, while revenue from managed service provider (MSP) operations, which support customers’ cloud deployment and management, increased 17%. Samsung SDS attributed CSP growth to demand for Samsung Cloud Platform (SCP) and expanded GPU cloud services for public-sector and external customers. Financial-sector AI transformation projects and shipbuilding enterprise resource planning (ERP) projects supported MSP revenue growth.

Samsung SDS is also expanding its AI computing services. It introduced an NVIDIA B300-based GPU cloud service in March and launched a neural processing unit (NPU) cloud service using FuriosaAI’s RNGD, pronounced “Renegade,” in July. These services provide GPU and NPU computing resources to meet customers’ AI requirements. They reflect the company’s existing business structure, which combines infrastructure provision with enterprise system implementation and operations.

In global infrastructure, Samsung SDS joined a collective investment by Samsung affiliates. On September 29, Samsung Electronics, Samsung C&T, Samsung SDS, Samsung SDI, Samsung Life Insurance and Samsung Fire & Marine Insurance announced an agreement to invest a combined $1 billion in Helix Digital Infrastructure. Samsung Electronics will contribute $500 million, with the other five companies contributing a combined $500 million. The announcement did not disclose Samsung SDS’s individual investment amount.

Founded by KKR (Kohlberg Kravis Roberts), Helix focuses on large-scale data centers, power generation and transmission, and fiber networks. The Samsung affiliates plan to explore cooperation drawing on their capabilities in semiconductors, cooling, construction, data center operations and power systems. Samsung SDS identified data center design, construction and operations, along with GPU cloud services, as its relevant capabilities. The announcement outlined the investment decision and intended cooperation but did not disclose Helix-related contracts or revenue for Samsung SDS.

Extending Manufacturing Systems Experience to Robotics Operations

In manufacturing, Samsung SDS is developing an integrated operations platform that connects production equipment with robots. At last month’s REAL Summit, the company announced its entry into the RX business and plans to introduce its robot orchestration platform from 2027. Orchestration means connecting multiple robots and pieces of equipment to coordinate tasks and manage operations. Samsung SDS is testing ways to automate manual factory tasks using robotics and physical AI.

The initiative builds on the company’s manufacturing systems experience. Samsung SDS said it has implemented manufacturing execution systems (MES) and equipment and logistics automation for approximately 430 customers over the past 25 years, spanning semiconductors, batteries, biotechnology, automotive, materials, components, and food and beverages. It is using that experience to develop robot-based manufacturing automation and technologies for managing production processes. The customer figure refers to its established manufacturing systems business, rather than adoption of the new robotics platform.

Samsung SDS is also working with physical AI companies. It announced a strategic investment in U.S.-based Walden Robotics through Samsung Venture Investment in June. Last month, it introduced its strategic partnership with Walden Robotics as part of its efforts to automate manual manufacturing tasks. The partnership focuses on connecting robotics technology with existing manufacturing systems.

A notable feature of Samsung SDS’s strategy is its extension of established capabilities into AI applications. In cloud services, computing resources and customer system operations are generating revenue, while in manufacturing, MES and automation experience underpin its work on integrated robotics operations. The Helix investment represents a separate initiative in global AI infrastructure cooperation. Although the three areas are at different stages, the strategy consistently combines infrastructure, system implementation and operational expertise.

The results reported so far include cloud revenue growth and the launch of AI computing services, while the robotics platform remains under development and validation. Samsung SDS has also outlined how it intends to apply its manufacturing experience to the new platform. Future performance will be measured by actual platform adoption, operating results and improvements in customer productivity. Its stated direction is to connect data center computing resources with practical AI use in enterprise workflows and manufacturing operations.

Read the Korean version

By Manripungchwi Jung Yeon-tae

October 9, 2026, marks the 580th anniversary of Hangeul Day, a national holiday celebrating the creation and promulgation of Korea’s unique writing system. This year carries special historical significance. It marks 580 years since King Sejong promulgated Hunminjeongeum, the original name of Hangeul, in 1446. It also marks exactly 100 years since our ancestors established the first celebration of Hangeul Day, originally called Gagyanal, in 1926 during the Japanese colonial period.

Five hundred and eighty years ago, King Sejong gave his people a writing system of their own. A century ago, our ancestors struggled to preserve that writing system while their country was under foreign rule. The letters we read and write so effortlessly today embody both the compassion of a king who cared deeply for his people and the dedication of those who fought to preserve their language and national identity. Yet how deeply do we truly appreciate this extraordinary legacy?

Hangeul is far more than a means of communication. It is a writing system born from compassion for humanity and a remarkable intellectual achievement that continues to demonstrate its value in the digital age and the emerging era of artificial intelligence.

1. King Sejong: A Visionary Leader Who Created an Alphabet for His People

In 1443, King Sejong created the 28 original letters of Hunminjeongeum, which were officially promulgated in 1446. At the time, Korea’s intellectual and administrative systems relied heavily on Chinese characters. For ordinary people who lacked access to formal education, learning these characters was extremely difficult. Many could neither express their thoughts in writing nor easily communicate their grievances to the authorities.

King Sejong was deeply troubled by this reality. His compassion is clearly reflected in the opening words of the Hunminjeongeum preface:

“Because the language of our country differs from that of China, it does not readily correspond to Chinese characters…”

King Sejong explained that he had created a new writing system so that ordinary people could easily learn it and conveniently use it in their daily lives. Behind these simple words lies an extraordinary philosophy of leadership. King Sejong possessed the highest political authority in Joseon. The existing system, in which literacy was concentrated among educated elites, largely benefited the ruling class.

Knowledge was power, and access to written knowledge helped preserve social privilege. Yet Sejong sought to lower those barriers. His guiding principle was that knowledge should not remain the exclusive privilege of a few but should become accessible to all. From a modern perspective, this was a revolutionary vision of broader access to knowledge and education.

King Sejong recognized the difficulties faced by ordinary people, studied their causes, developed an innovative solution, and made that solution available to society. His leadership combined compassion, scientific inquiry, innovation, and a profound commitment to the public good. He did not create Hangeul to glorify his own authority. He created it so that ordinary people, whose names might never appear in history books, could express their thoughts and communicate their feelings.

To me, this is the essence of King Sejong’s greatness.

2. Hangeul: A Remarkable Invention Combining Science and Philosophy

Hangeul occupies a distinctive position in the history of writing systems. Unlike many scripts that evolved gradually over centuries, Hangeul has a historically documented creator, purpose, date of creation, and underlying design principles. Its basic consonant letters—ㄱ, ㄴ, ㅁ, ㅅ, and ㅇ—were designed to reflect the shapes or characteristics of the human speech organs involved in producing sounds. Its original vowel system incorporated the philosophical concept of Cheon-Ji-In (天地人), representing Heaven, Earth, and Humanity.

Thus, the scientific analysis of human speech and the philosophical understanding of nature and human existence were integrated into a single writing system. Another remarkable characteristic of Hangeul is its systematic combination of a relatively small number of letters into a vast range of syllables. Modern Hangeul uses 24 basic letters—14 consonants and 10 vowels. Through systematic combinations of initial consonants, vowels, and optional final consonants, the Unicode standard accommodates 11,172 precomposed modern Korean syllables.

This demonstrates how the structural regularity of Hangeul can be represented systematically in digital technology. Of course, King Sejong could not have anticipated computers or artificial intelligence nearly six centuries into the future. Nevertheless, it is remarkable that a writing system created in the 15th century could be adapted so effectively to modern information technology.

The greatness of Hangeul lies not merely in its relative ease of learning. Its true brilliance lies in the scientific analysis of human speech and the ability to express complex sounds through a concise, systematic set of principles. This intellectual achievement deserves recognition not simply as a source of Korean national pride but as an important contribution to human civilization.

3. How Hangeul Helped Transform Korea’s Destiny

The promulgation of Hangeul did not immediately make literacy universal. For centuries, Chinese characters remained dominant in scholarly and official communication, while Hangeul gradually spread through everyday life. It appeared in personal letters, literary works, songs, and religious texts, enabling a wider range of people, including women with limited access to classical education, to express themselves in writing. Over time, Hangeul became an essential means of preserving and transmitting Korean culture.

Its importance became especially apparent during one of the darkest periods of Korean history. In 1926, while Korea was under Japanese colonial rule, Korean language scholars and cultural activists established Gagyanal, the predecessor of Hangeul Day. For these individuals, preserving the Korean language and writing system meant more than protecting a method of communication. It meant protecting the identity, memory, and cultural continuity of a nation that had lost its sovereignty.

Scholars and educators such as Ju Si-gyeong and Choe Hyeon-bae devoted their lives to studying, preserving, and developing the Korean language. Many language activists endured hardship and repression in their efforts to safeguard Korea’s linguistic heritage. Their dedication helped ensure that Hangeul would survive and flourish.

Following Korea’s liberation in 1945, Hangeul played a significant role in expanding education, literacy, public communication, and cultural development. Korea’s educational achievements and economic development cannot be attributed to Hangeul alone. Educational policies, social investment, industrialization, and the commitment of the Korean people were also essential. Nevertheless, Hangeul’s systematic and accessible structure provided an important foundation for literacy and the dissemination of knowledge.

King Sejong created Hangeul, but generations of Koreans preserved, developed, and passed it down to us. The freedom to read, write, learn, and communicate in our own language is not something we should take for granted. It is a precious inheritance made possible by the wisdom and sacrifices of our ancestors.

4. Hangeul in the Digital and Artificial Intelligence Eras

The emergence of computers and the internet in the late 20th century transformed the nature of written communication. Letters that had once existed primarily on paper became digital information. Hangeul’s systematic principles of combining consonants and vowels proved useful in the development of Korean keyboard input methods and digital character-processing technologies. Today, Koreans can easily type Hangeul on smartphones, search for information online, create digital documents, and communicate instantly with people around the world.

We have now entered another period of profound technological transformation: the age of artificial intelligence. Generative AI systems, including ChatGPT, can process human language, generate text, analyze information, and assist with increasingly complex intellectual tasks. In the past, people had to learn specialized programming languages and commands to interact with computers. Today, computers can increasingly respond to instructions expressed in ordinary human language.

This development creates new opportunities for Hangeul and the Korean language. For AI to understand and use Korean accurately, it requires extensive, reliable, and diverse Korean-language data. Knowledge concerning Korean history, philosophy, science, technology, law, administration, literature, and culture must be systematically preserved and made accessible in digital form. At the same time, AI must learn to handle the distinctive grammatical and semantic features of Korean, including particles, verb endings, honorific expressions, contextual meaning, and the frequent omission of sentence elements.

An important distinction must be made here. Hangeul is a writing system, while Korean is a language. The effectiveness of AI in processing Korean depends not simply on the scientific structure of Hangeul but also on linguistic complexity, data quality, computational methods, and model architecture. Nevertheless, Hangeul’s systematic structure remains a valuable foundation for digital text representation and processing.

I believe that a nation’s linguistic capabilities will become an increasingly important component of its digital sovereignty and knowledge sovereignty. The ability to create, preserve, analyze, and utilize knowledge in one’s own language will become a significant element of national competitiveness.

Nearly six centuries ago, King Sejong was concerned that ordinary people could not fully express themselves because they lacked access to a suitable writing system. Today, we should similarly be concerned about those who cannot fully participate in society because they lack access to digital technologies and artificial intelligence. The compassion that inspired the creation of Hangeul can therefore guide our responsibilities in the AI era. Technology should not remain the privilege of a small, highly educated group.

Its benefits should be accessible to everyone. In this sense, King Sejong’s philosophy remains profoundly relevant.

5. What We Must Pass Down to Future Generations

Today, interest in the Korean language continues to expand around the world. The global popularity of K-pop, Korean dramas, films, literature, cuisine, and other aspects of Korean culture has inspired people from many countries to study Korean. The language that Koreans once struggled to preserve under colonial rule has become one that people around the world voluntarily choose to learn. What an extraordinary historical transformation.

Yet the globalization of Hangeul should mean more than simply increasing the number of people who can read and write Korean. It should also provide opportunities to share the intellectual and cultural values that shaped its creation. King Sejong’s compassion for ordinary people, his respect for human dignity, his commitment to education, and his dedication to scientific inquiry remain relevant to people of every nationality.

At the same time, we must remember that genuine cultural pride does not require us to diminish other languages or cultures. Every language and writing system embodies its own history, identity, and cultural significance. True confidence comes from understanding the value of our heritage while respecting the heritage of others.

There is another responsibility we must recognize. In contemporary Korean society, unnecessarily complicated terminology and the excessive use of foreign expressions can sometimes make communication difficult. Even official documents intended for the public are occasionally filled with language that ordinary citizens struggle to understand. What would King Sejong think if he saw such practices today?

He created Hangeul so that people could communicate more easily. To honor his legacy, we should strive to make public language clear, accurate, and accessible. Loving Hangeul does not mean rejecting foreign words or resisting linguistic change. It means preserving the ability of our language to communicate meaning effectively while adapting to new knowledge and cultural developments.

Above all, we must teach our children and grandchildren more than the letters themselves. They should understand why Hangeul was created, what values inspired its creation, and how previous generations struggled to preserve it. They should learn the importance of compassion for the disadvantaged, equal access to education, scientific curiosity, innovation, and responsible leadership.

The true purpose of historical education is not merely to celebrate past achievements but to apply the wisdom of our ancestors to the challenges of the present. That is the inheritance we should pass on to future generations.

6. King Sejong’s Wisdom: Illuminating the Next Thousand Years

On this Hangeul Day, I find myself reflecting once again on King Sejong. Why do we remember him as one of the greatest leaders in Korean history? It is not because he conquered vast territories or constructed magnificent palaces. His greatness lies in something far more enduring.

He created a writing system that enabled ordinary people to express their thoughts and feelings. That writing system gradually transformed countless lives. It allowed mothers to write letters to their children, loved ones to communicate their affection, and ordinary people to preserve their experiences and memories. It helped open the world of learning to people who might otherwise have remained excluded from written knowledge.

Today, Koreans use Hangeul to conduct scientific research, create literature, communicate globally, and interact with artificial intelligence. Perhaps King Sejong’s greatest legacy was not simply the writing system itself. It was his belief that ordinary people should have the opportunity to become active participants in the world of knowledge.

As I reflect on this history, I am reminded that the educational, cultural, and technological achievements we enjoy today did not emerge by accident. They rest upon the accumulated wisdom, sacrifice, and dedication of countless generations. We owe our ancestors profound gratitude and respect. But gratitude must lead to action.

We must preserve the cultural heritage they entrusted to us, develop new knowledge and technologies, and leave future generations a richer intellectual and spiritual inheritance. As artificial intelligence transforms our world, we should ask ourselves two fundamental questions. For whom should technology exist? Whom should knowledge serve?

Nearly 580 years ago, King Sejong offered an enduring answer.

Technology and knowledge should serve the people.

The greatness of technology lies not only in the complexity of the problems it can solve but also in the number of human lives it can improve. This principle connects King Sejong’s vision with the responsibilities of our own technological age.

On the 580th anniversary of Hangeul’s promulgation, I offer my deepest respect and gratitude to King Sejong. I also bow in appreciation to the countless scholars, educators, patriots, and ordinary citizens who protected and developed our language throughout history. With the writing system they preserved, we continue to record our thoughts, share our love, explore the universe, and imagine tomorrow’s possibilities.

Our responsibility is clear. We must do more than take pride in King Sejong’s extraordinary legacy. We must carry his spirit forward by building a society that future generations will be proud to inherit.

I hope that one day our descendants will look back upon our era and say:

“Our ancestors not only preserved the writing system created by King Sejong but also transformed its wisdom into a new intellectual foundation for the age of artificial intelligence.”

If we can accomplish that, the seed of wisdom planted nearly six centuries ago will continue to flourish, illuminating human civilization for another thousand years. Hangeul is a proud inheritance from our history. King Sejong’s compassion for humanity is an enduring legacy we must carry into the future.

October 9, 2026
Commemorating the 580th Anniversary of Hangeul Day

By Manripungchwi Jung Yeon-tae

About the Author

Jung Yeon-Tae, who writes under the pen name Manri Pungchwi, is a cultural thinker and policy strategist. His research and writing connect history, philosophy, science and technology, interpreting Korea’s traditional culture as a source of future value in an era of civilizational change and artificial intelligence.

Read the Korean version

Series Order

Part 1 | When a Company Fails to Pay for a Transaction
Part 2 | Continued Use of a Business Name and the Transferee’s Liability
Part 3 | A Company Director’s Liability for Damages
Part 4 | The Person Hiding Behind a Company
Part 5 | What to Consider After Being Assaulted
Part 6 | What Is an Appropriate Criminal Settlement Amount?
Part 7 | If You Cannot Remember What Happened After Drinking, Could It Constitute Quasi-Rape?
Part 8 | Freelancers and Severance Pay
Part 9 | What to Consider When Assigning a Claim to Secure Business Financing
Part 10 | Can You Divorce a Foreign Spouse Who Has Disappeared?
Part 11 | Can a Tenant Terminate a Lease If the Landlord Delays Repairs?
Part 12 | Can a Subcontractor Receive Payment If the Main Contractor Disappears?
Part 13 | What to Know Before Your First Police Interview
Part 14 | Can a Buyer Recover a Deposit Before Receiving Approval for a Contract in a Land Transaction Permit Zone?
Part 15 | Is Entering Another Person’s Home While Concealing a Criminal Purpose Trespassing?
Part 16 | Impersonating Another Person on Instagram Can Lead to Criminal Punishment
Part 17 | When Does Indecent Assault Occur?
Part 18 | When Repairs Are Required After a Water Leak, Can the Tenant Decide How the Work Is Done?
Part 19 | How Far Does a Real Estate Agent’s Duty to Explain Extend?
Part 20 | Does Swearing in a Residential Area Constitute Criminal Insult?
Part 21 | Can Damages Be Claimed for Incorrect Tax Advice Without a Formal Engagement Agreement?
Part 22 | Can Families Claim Damages for the Exhumation of an Ancestral Grave?
Part 23 | How Far Does Individual Liability Extend in Cambodia-Based Scam Rings?
Part 24 | When Does Filming Sexual Activity Between Partners Constitute a Crime?
Part 25 | Childcare Corporation Assets Can Pass to Designated Recipients After Dissolution
Part 26 | What Buyers Must Check Before Purchasing Land with Graves

By Hong Seung-kwon, Principal Attorney, PAN Law Office (skh@law-pan.com)

A prospective buyer of land or forest property may discover that several old graves are located within its boundaries. When the graves appear to have been neglected for many years, it can be tempting to assume that they are unclaimed and can simply be relocated after the purchase. Land containing graves, however, requires the buyer to examine legal issues involving customary gravesite rights and statutory exhumation procedures in addition to ownership.

An unclaimed grave does not simply mean a grave that has not been maintained. Even if the vegetation has not been cleared for years or the headstone bears no name, a person connected to the grave may still exist. Buyers should not classify a grave as unclaimed solely from its appearance. They must investigate when and how it was established and whether any relatives or other connected persons can be identified.

Particular attention should be paid to the customary gravesite right known in Korean law as bunmyogijigwon. This customary-law right allows a person to use another person’s land to protect and maintain a grave located there. For graves installed without the landowner’s consent before the Act on Funeral Services, Etc. took effect on January 13, 2001, the Supreme Court recognizes acquisition of this right by prescription when the land was occupied peacefully and openly for 20 years. Graves installed without the landowner’s consent after that date generally cannot acquire the right through prescription. A separate review remains necessary because some graves may have been established with the landowner’s consent.

How, then, can a grave be relocated when no connected person can be identified? Under Article 27 of the Act on Funeral Services, Etc., a landowner may obtain an exhumation permit from the competent mayor, county governor or district head after completing the prescribed procedures for a grave installed without the owner’s consent. Notice of the intended exhumation must be given or published at least three months in advance. When the connected person is unknown, notice must be published at least twice in the manner prescribed by law, with the second notice issued at least 40 days after the first. If no one comes forward, the landowner may obtain a permit, cremate the remains and keep them enshrined for the legally prescribed period.

The situation becomes more complicated when several graves are located on the property. Each grave may have a different date of establishment, a different connected person and a different legal basis for using the land. They cannot all be treated collectively as unclaimed graves. Some may be exhumed through statutory procedures, while others may require negotiation with relatives over relocation. The process may generate substantial costs and delay a development project for an extended period.

Before signing a contract to purchase such land, the buyer should inspect the site to establish the number and location of the graves and, where possible, obtain information about when they were installed and who is connected to them. The sales contract should clearly identify who is responsible for relocating the graves and paying the associated costs. It should also include a special clause allowing the buyer to terminate the contract or claim damages if the seller fails to complete the relocation by a specified deadline. The parties may also consider making completion of the relocation a condition for payment of the remaining purchase price.

A person who has already purchased the land must not excavate a grave or move human remains without authorization merely because they own the property. Excavating a grave without following the lawful procedures may result in criminal liability. A transaction involving land with multiple unclaimed graves therefore requires an assessment of the feasibility, time and cost of relocation as well as the purchase price. Because customary gravesite rights do not appear in the land registry, buyers must inspect the property and obtain relevant information directly from the seller.

About the Author

Hong Seung-kwon is the principal attorney of PAN Law Office. His practice covers civil and criminal litigation as well as emerging legal issues in the digital era. He continues to examine how technological innovation, property rights and changes in regulatory policy affect law and market order.

Read the Korean version

The European Union is negotiating to reduce imports of Chinese-made vehicles while simultaneously seeking China's cooperation in easing rare earth export procedures. Europe wants to protect its automotive industry from the growing market share of Chinese manufacturers, yet it remains heavily dependent on Chinese supply chains for critical materials used in vehicle production. The contradiction illustrates a central challenge in European industrial policy: China is both a major competitor in the automotive market and an indispensable supplier of manufacturing inputs.

According to Reuters and the Associated Press on October 9, the EU and China reached a preliminary understanding following two days of trade negotiations in Beijing. The discussions covered a reduction in Chinese hybrid and plug-in hybrid vehicle exports to Europe and measures to facilitate export licensing for rare earths and permanent magnets. The EU also sought improved access to the Chinese market for European products, including automotive components, olive oil and footwear. However, the specific mechanisms and legal status of the proposed export restrictions have yet to be finalized, and further discussions among EU leaders are expected.

European Commissioner for Trade and Economic Security Maroš Šefčovič said the negotiations had opened the possibility of reducing Chinese hybrid vehicle exports to the EU by more than half. He described the outcome as "a crucial first step – but only a first step." His remarks underscored the need for continued negotiations on automotive trade, bilateral trade imbalances, market access and critical mineral supplies.

Surging Chinese Vehicle Imports Put Pressure on Europe's Automotive Industry

The EU's demand for reduced Chinese vehicle exports reflects mounting pressure on the competitiveness of European automakers. Chinese manufacturers have expanded their presence in Europe through large-scale production and competitive pricing. European carmakers, meanwhile, face both U.S. tariff pressures and intensifying competition from lower-priced Chinese vehicles, prompting adjustments to manufacturing operations and employment.

According to Reuters, EU imports of plug-in hybrid vehicles increased by 86% in the 12 months through September, while import prices fell by approximately 20%. China accounted for more than half of those imports. The figures illustrate how Chinese manufacturers have expanded their European sales while intensifying price competition.

The EU introduced additional tariffs on Chinese-made battery electric vehicles in 2024 following an anti-subsidy investigation. Chinese manufacturers, however, have not limited their European expansion to fully electric vehicles. Their growing exports of hybrid and plug-in hybrid models have complicated the EU's efforts to manage competition through existing trade defense measures.

The EU's push to reduce Chinese hybrid vehicle exports is a response to this shift. Šefčovič said the proposed arrangements could prevent millions of Chinese vehicles from entering the European market over the next four years. That remains a projected outcome of the negotiations, however, rather than evidence that exports have already fallen by half or that binding volume restrictions have taken effect.

Import restrictions alone will not necessarily restore the competitiveness of Europe's automotive industry. Reduced inflows of Chinese vehicles could ease pricing pressure on European manufacturers. But the benefits of market protection may remain limited unless automakers also improve production efficiency, technological development, electrification strategies and their ability to respond to changing consumer demand.

Higher vehicle prices could also shift part of the cost to European consumers. If the supply of lower-priced Chinese vehicles declines, domestic manufacturers may benefit from improved market conditions, while consumers could face fewer choices and weaker price competition. The economic benefits and costs of industrial protection would therefore be distributed differently between manufacturers and buyers.

Rare Earth Supply Chains Give China Additional Leverage

Europe's need to maintain cooperation with China, even as it seeks to restrict Chinese vehicle imports, stems partly from its dependence on rare earths and permanent magnets. Rare earth elements are used in electric vehicle motors, wind turbines, industrial robots, defense equipment and other advanced manufacturing applications. Certain high-performance permanent magnets are particularly important for improving the efficiency of electric and hybrid vehicle powertrains.

China holds a strong position not only in rare earth mining but also in processing, separation and permanent magnet manufacturing. Even if Europe succeeds in limiting China's influence in the finished-vehicle market, reducing its dependence on Chinese components and raw materials will take considerably longer. Vehicle manufacturing and critical material procurement therefore present European policymakers with different, and sometimes competing, objectives.

During the negotiations, China indicated that it would continue supporting a so-called green channel to facilitate export licensing for rare earths and permanent magnets. Such arrangements could reduce administrative uncertainty for European manufacturers seeking essential materials. However, faster licensing procedures would neither eliminate China's authority over export controls nor resolve Europe's underlying supply chain dependence.

According to China's Ministry of Commerce, Commerce Minister Wang Wentao argued that China was not the fundamental cause of the economic difficulties facing the EU and should instead be regarded as a partner in addressing them. His remarks reflected Beijing's position that Europe's industrial challenges cannot be attributed solely to rising Chinese imports.

Indeed, Europe's automotive difficulties extend beyond competition from Chinese manufacturers. High production costs, substantial investments in electrification, changing consumer preferences and U.S. trade policies have all contributed to pressure on the industry. Restrictions on Chinese imports may ease some competitive pressures, but they are unlikely to resolve these broader structural challenges on their own.

China also has economic reasons to preserve stable trade relations with Europe. As trade barriers rise in the United States and other major markets, Europe remains an important export destination for Chinese manufacturers. Beijing may therefore see value in accepting some adjustments to automotive trade in exchange for greater market access in other sectors and more predictable bilateral commercial relations.

Automotive Trade and Rare Earths Become Linked Negotiating Issues

The negotiations demonstrate how vehicle exports and critical mineral supplies can become interconnected elements of trade diplomacy rather than separate commercial disputes. The EU wants to reduce imports of Chinese vehicles while securing reliable access to Chinese rare earths. China, meanwhile, is discussing potential constraints on automotive exports while retaining its established position in critical material supply chains.

The EU's trade deficit with China expanded to an average of more than €1 billion per day last year. Beyond adjustments to automotive exports, the two sides discussed improving access to the Chinese market for approximately €4 billion worth of European products. The discussions covered automotive components, olive oil and footwear, along with possible tariff reductions on certain goods.

The scope of the negotiations shows that the dispute extends well beyond the automotive sector. Europe is pursuing industrial protection and export expansion at the same time, while China is addressing vehicle trade, critical mineral supplies and imports of European consumer goods within a broader negotiating framework. How EU member states with differing industrial interests respond to the proposed arrangements will also be an important factor.

Implications for South Korea's Automotive and Battery Industries

The preliminary understanding could reshape competitive conditions for South Korean automakers and battery manufacturers. If Chinese hybrid vehicle exports to Europe are substantially reduced, Korean automakers may gain opportunities in certain market segments. However, if the EU expands policies favoring European production and local sourcing, vehicles and components exported from South Korea could face additional challenges.

The implications for the battery industry cannot be assessed solely through changes in vehicle sales. Europe's electrification policies, local manufacturing requirements and the availability of Chinese battery materials will also influence industry conditions. Rare earths and permanent magnets belong to supply chains distinct from those of battery raw materials, but they remain essential components of vehicle electrification and are therefore relevant to broader procurement strategies.

South Korean companies will need to assess not only the possibility of reduced Chinese vehicle exports but also the expansion of European manufacturing, China's export licensing policies for critical materials and changes in European consumer demand. Restrictions on Chinese vehicles should not automatically be interpreted as a corresponding increase in sales of Korean-made vehicles.

The practical significance of the negotiations will become clearer during implementation. Key indicators will include how Chinese hybrid vehicle export volumes are managed, whether rare earth export licenses are processed more efficiently and whether European products gain meaningful access to the Chinese market.

Europe's continued reliance on Chinese rare earth supplies, even as it seeks to restrict Chinese vehicle imports, exposes the limits of industrial protection. Trade barriers may reduce competitive pressure in a particular market, but they cannot provide the supply chain independence that Europe has yet to establish in raw materials and components. The long-term competitiveness of Europe's automotive industry will depend not only on how effectively it limits Chinese vehicle imports but also on whether it can diversify critical material supplies and improve productivity and technological capabilities.

The same considerations apply to South Korea. Korean manufacturers need to capitalize on market opportunities created by trade protection while reducing excessive dependence on individual countries for essential materials and components. They must also adapt to changes in European manufacturing and supply chains. The EU-China negotiations illustrate how industrial competitiveness increasingly depends on a country's ability to manage competition in finished products alongside access to the critical materials required to produce them.

Read the Korean version

SK Group Chairman Chey Tae-won visited a proposed semiconductor complex site in Jeonnam-Gwangju Integrated Special City on October 9, pledging to accelerate development of a semiconductor cluster in the Honam region of southwestern Korea. He said semiconductor demand was rising faster than expected and that bringing forward the full completion of the Yongin semiconductor cluster by 12 years might still be insufficient. Chey raised the possibility of building fabs simultaneously in Yongin and Jeonnam-Gwangju as he assessed how quickly production facilities could be constructed at the new site. He said a specific groundbreaking date would depend on a coordinated review of site preparation, electricity and water supply plans, and construction schedules.

Chey toured the Gwangju military airport site and a nearby ammunition depot identified as a key candidate site for preparation, accompanied by SK hynix CEO Kwak Noh-jung and other senior executives. He met Min Hyung-bae, mayor of Jeonnam-Gwangju Integrated Special City, and other officials for a briefing on the industrial complex development plan and preparations for land, power, water and transport infrastructure. He then instructed executives to “work closely with the central and local governments to advance the southwestern semiconductor cluster as quickly as possible.”

At recent internal and external meetings, Chey has stressed the urgency of expanding production capacity, saying, “To lead in the AI era, we need to more than double memory semiconductor production capacity within the next five years.” During the visit, he reiterated that the pace of expansion must increase to keep up with demand. “Semiconductor demand is growing faster than expected, requiring us to accelerate the expansion of production facilities, and even though we have brought forward the overall construction schedule for the Yongin semiconductor cluster by 12 years, we expect that may still be insufficient,” he said. “We may need to build factories in Yongin and Jeonnam-Gwangju simultaneously.”

Full Yongin Completion Brought Forward by 12 Years, Honam Plan About Two-Thirds Its Size

During an on-site question-and-answer session, Chey explained that he had come because seeing the location firsthand was important. He said semiconductor demand was expected to continue growing rapidly, requiring a corresponding increase in the pace of production capacity expansion. He clarified that the 12-year acceleration referred to completing the development of the entire Yongin site, rather than opening a single fab. Even that accelerated plan might not meet demand, potentially requiring additional sites and simultaneous construction.

Chey said the facilities currently envisaged for Jeonnam-Gwangju would be about two-thirds the size of the Yongin cluster. He did not specify the number of fabs. He added that the Honam cluster could grow further if more land became available and demand increased. “This visit is to see firsthand how quickly we can build factories in Jeonnam-Gwangju,” he said, adding, “We will build as quickly as possible.”

Construction Timing Under Review as Power and Water Plans Are Coordinated

Chey said it was too early to give a definitive construction start date. Further calculations were needed to determine how local preparations and national water and electricity supply plans would align with construction. He suggested that related plans could be announced soon. Asked whether work could begin before the military airport was relocated, he said construction would proceed as quickly as possible once conditions allowed, while noting that the precise timing required further study.

Chey also raised the possibility of beginning construction before electricity and water infrastructure was ready. His remarks indicated that, if conditions permitted a start and the piling schedule could be coordinated, construction could proceed even if utility provision was delayed. He said saving on construction costs at the site was not a particular priority. The comments underscored his intention to secure additional capacity as quickly as possible to keep pace with demand.

“On behalf of the special city’s 3.2 million residents, I thank Chairman Chey Tae-won for clearly demonstrating his commitment to advancing the semiconductor cluster,” Min said. “We will firmly support not only essential infrastructure such as land, electricity and water, but also talent development and living conditions, so that SK hynix can reliably build and operate world-class semiconductor factories in an optimal environment.” The local government pledged to support workforce availability and residential conditions alongside the infrastructure needed for construction and operation.

SK hynix unveiled plans in June for a new KRW 400 trillion semiconductor cluster in the Honam region. Since then, executives have visited the Gwangju site about 10 times to hold discussions with the local government. Chey’s visit focused on inspecting the proposed site and infrastructure preparations, while setting a direction for faster implementation through cooperation with central and local authorities.

“Surging global demand for AI memory semiconductors is accelerating market growth, while competition is also becoming increasingly intense,” an SK hynix representative said. “With strong support from the central and local governments, we will continue investing at pace to help strengthen Korea’s leadership in AI semiconductors.”

Read the Korean version

Anyflo announced its emergence from stealth and the acquisition of blockchain developer Native in Singapore on October 8, outlining plans to simplify stablecoin payments for enterprises. Incubated at Mysten Labs, an initial contributor to the Sui blockchain, the startup is developing a platform to coordinate payments across multiple blockchain networks. Mysten Labs has also invested $1 million in the company. Financial terms of the Native acquisition were not disclosed.

Founded and led by Lola Oyelayo-Pearson, formerly general manager of product at Mysten Labs, Anyflo aims to manage the blockchain connections, liquidity and transaction routing needed for stablecoin payments. Its target customers include marketplaces, neobanks and payroll providers integrating payments into their own products. The platform is intended to let these businesses access stablecoins across networks without building and managing each connection themselves.

The acquisition brings Native’s engineering team and cross-chain intellectual property into Anyflo. According to the announcement, Native raised approximately $2 million before the acquisition and developed a Bitcoin lending platform on Sui. It also ranked among the top five of 630 projects in Sui Overflow 2025. Native co-founder Robert Zaremba and his team will contribute their blockchain infrastructure expertise to Anyflo’s development.

Oyelayo-Pearson, whose product background includes Shopify Money and Capital One, said businesses are primarily seeking lower money-transfer costs, stronger margins and better customer experiences. She described Anyflo’s role as making blockchain-based payment infrastructure useful to those business objectives while reducing the technical complexity involved.

Anyflo plans to begin private alpha testing in October, followed by a production preview and the onboarding of liquidity providers. A beta release is scheduled for the first quarter of 2027. The announcement did not provide customer adoption figures or measured results for payment costs and settlement times.

Mysten Labs President Kevin Boon said Oyelayo-Pearson and the Native team are addressing the practical challenge of enabling businesses to use blockchain-based payments without becoming blockchain specialists. Mysten’s support combines incubation and investment with technical expertise from its work on Sui.

Anyflo’s planned service focuses on the infrastructure connecting enterprises to stablecoin payment networks. As development progresses, its ability to translate that integration into measurable improvements in cost, reliability and payment processing will be central to assessing its business value.

Read the Korean version

The circumstances differ. OpenAI's reported revenue gap reflects differences in reporting methodology rather than evidence of a decline in actual sales. Firmus, by contrast, struggled to secure sufficient investor support for a proposed valuation that depended heavily on future expansion. Both cases demonstrate why headline growth figures and ambitious business projections require closer examination before they can support multibillion-dollar valuations.

OpenAI's $20 Billion Revenue Gap Reflects Different Reporting Methods

The Financial Times reported on October 8, citing financial information provided to investors, that OpenAI's annualized revenue was approaching $50 billion as of the end of September. That was approximately 28.6% below the $70 billion figure reported by some media outlets the previous month. Reuters subsequently confirmed the approximately $50 billion figure with a person familiar with the matter. OpenAI did not respond to Reuters' request for comment.

The discrepancy largely concerns how OpenAI and its competitor Anthropic account for revenue generated through cloud partners. Anthropic reportedly includes sales through Amazon Web Services and Google Cloud in its annualized revenue calculations, while OpenAI does not account for comparable sales in the same way. The difference became apparent as investors sought to compare the companies' growth on a consistent basis.

Annualized revenue run rate estimates a company's yearly sales based on revenue generated over a shorter period, often by multiplying the most recent month's revenue by 12. It provides a useful indication of a rapidly growing company's current business scale, but it should not be confused with revenue recognized over a completed financial year or with cash actually collected.

The difference between the two reported figures does not mean OpenAI's revenue fell by $20 billion. Because the figures reflect different reporting scopes, they cannot establish whether the company's growth rate has accelerated or slowed. A reliable assessment would require comparable period-by-period revenue figures calculated under consistent definitions, alongside recognized revenue, operating expenses and cash flow.

The distinction matters particularly when investors value AI companies using revenue multiples. Including or excluding sales through cloud partners can materially change the apparent size of a business. Even rapid revenue growth does not necessarily translate into profitability once model training, inference and computing costs are taken into account.

Firmus' IPO Withdrawal Exposes the Challenge of Valuing Future Capacity

While OpenAI's revenue reporting attracted scrutiny, another test of AI valuations emerged in Australia. Firmus, an AI data center company backed by NVIDIA, Coatue Management and Blackstone, withdrew its planned $5 billion IPO on October 9 and said it would pursue financing in private markets. The offering would have been Australia's second-largest IPO.

Firmus had proposed an offer price of A$11 per share, implying an equity valuation of approximately A$30.6 billion. That was nearly three times the A$10.5 billion valuation assigned to the company during a fundraising round in early August.

The company develops modular AI data centers using proprietary energy and cooling technology. It currently operates two leased data center facilities in Melbourne and Singapore, while five additional projects across the Asia-Pacific region remain in early development.

Firmus had outlined an ambition to generate $5 billion in annual earnings from its data center business within five years. Investors, however, questioned whether its existing operating record could support the proposed valuation and the scale of its planned expansion. The rapid increase in valuation, combined with a relatively limited track record in large-scale data center development, added to those concerns.

Joseph Koh, a portfolio manager at Blackwattle Investment Partners, which reviewed the offering but did not participate, told Reuters that the proposed price appeared to assume near-perfect execution of the company's future plans. He indicated that investors had insufficient grounds at this stage to be confident that those projections would be achieved.

Firmus said market volatility and the proposed offering terms did not adequately reflect its long-term growth prospects, prompting the decision to withdraw the IPO. The outcome also reflected company-specific concerns, including valuation, financing requirements and execution risk. It should therefore not be interpreted as a broad rejection of AI data center investment.

Nevertheless, the episode illustrates the limits of assigning present-day valuations to infrastructure capacity and earnings that remain largely prospective. Announced projects and long-term customer agreements do not automatically translate into operating revenue. Actual profitability depends on construction completion, utilization rates, customer demand, power and cooling expenses, and the ability to service debt.

The OpenAI and Firmus cases expose different weaknesses in the way AI businesses can be valued. For OpenAI, the central issue is how existing revenue is defined and measured. For Firmus, it is how much of a future business plan investors are prepared to recognize in today's equity valuation.

AI model developers frequently emphasize customer growth and annualized revenue as evidence of commercial momentum. Yet evaluating their profitability also requires an understanding of inference costs, cloud infrastructure expenses and customer acquisition costs. For data center operators, contracted capacity and projected revenue must be assessed against actual utilization, operating expenses, depreciation and financing costs. Large debt-funded infrastructure projects may require years of sustained operation before their investments generate adequate returns.

Neither development provides sufficient evidence that investor appetite for AI has broadly weakened. They do, however, demonstrate that different revenue definitions can produce materially different growth figures, while aggressive valuations based on future infrastructure plans may fail to secure investor support.

The financial value of an AI company cannot be established by announced investment commitments or contract volumes alone. Investors need to know how revenue is calculated, whether planned facilities are operating, how much customers actually pay and how much cash remains after the costs of delivering the service. OpenAI and Firmus offer two reminders that the economic value of AI growth ultimately depends on financial results that can be consistently measured and independently assessed.

Read the Korean version

Sumsub launched its Reusable KYC Gateway in Singapore on October 7 at TOKEN2049, with Opera’s self-custodial stablecoin wallet MiniPay as its first live integrator. The service allows users to reuse a verified identity when accessing regulated services within MiniPay’s partner network, reducing the need for repeated document uploads and selfie checks.

According to Sumsub, MiniPay partners adopting the approach have seen verification conversion increase from approximately 70% to 95%, while KYC completion time has fallen by up to 50%. The announcement did not specify the measurement period or sample size behind those figures.

MiniPay operates in more than 70 countries and has recorded over 20 million wallet openings. Its partners, including fiat on- and off-ramp providers, card issuers and payment companies, have traditionally conducted separate know-your-customer checks. This can require users to submit the same identity documents each time they access a new service within the wallet’s ecosystem.

The gateway enables users to access an existing Sumsub ID or complete an initial verification using an identity document and a selfie. Sumsub said its platform has 15 million existing Sumsub IDs. Users then explicitly consent to link their verified identity to their MiniPay account.

When accessing a regulated partner for the first time, users consent to share their verified identity with that provider. Sumsub transfers the information directly to the partner, allowing MiniPay to facilitate access without collecting, storing or handling identity documents or biometric data.

Identity reuse does not remove each partner’s compliance responsibilities. Partners continue to apply their own policies and checks, with additional verification required for higher-risk transactions. Eight MiniPay partners already use Sumsub for verification, according to the announcement.

“Repeated KYC checks across the same ecosystem are inconvenient for everyone, and one in three applicants we see have already verified with us before,” said Andrew Novoselsky, Chief Product Officer at Sumsub. He said the gateway allows MiniPay users to reuse a verified identity across the partner network while controlling what they share and with whom.

Murray Spark, Senior Director of Business Development at Opera, said user choice over identity sharing complements MiniPay’s self-custodial model, in which users hold their own funds.

“They decide what to share and with whom, and that matters more and more as our partner network expands across new markets,” Spark said.

Reusable KYC Gateway joins Sumsub’s existing Reusable KYC product line, which also includes Reusable KYC Share. The company said the gateway is available to non-regulated businesses connecting users with regulated partners, including non-custodial wallets and decentralized trading platforms.

Series Overview

Part 1 | Why Does AI Get Things Wrong So Convincingly?
Part 2 | Why the OXMV Loop and Evidence Ledger Are Necessary
Part 3 | Mapping Knowledge for AI with Ontology
Part 4 | OpenAI to End ChatGPT’s Custom GPT Era on December 11, 2026
Part 5 | Taking Aim at ChatGPT’s Chronic “Hallucination” Problem
Part 6 | Can AI Detect Financial Cyberattacks Before They Escalate?
Part 7 | GPT-6.1 Sol: From Conversational AI to a Work Platform

By Manripungchwi Jung Yeon-tae

ChatGPT is changing again. This time, however, the change is not simply about moving from one model number to another. Nor is it merely a redesigned interface or a few additional menu items. What is happening now is closer to a structural reorganization of the entire AI ecosystem.

Functions that were once mixed together inside a single ChatGPT experience are being separated into more specialized layers, with each layer becoming more powerful. The most important question is therefore no longer simply, “How much smarter is the new model?” A better question is, “What role will GPT, Plugins, dot, Work, Page, and Space each play in the new ChatGPT environment?” Once that structure becomes clear, the recent changes start to make much more sense.

1. GPT-6.1 is not just “a smarter model”

The significance of GPT-6.1 lies not only in improved reasoning or technical performance. The larger change is how ChatGPT itself is evolving. ChatGPT is moving away from being a system that simply produces text answers and toward one that can choose the most appropriate interface, tool, workflow, or visual representation for the task. This is where the idea of an Intelligent UI becomes important.

Depending on the question, ChatGPT may present text, comparison views, diagrams, calculators, interactive tools, visual summaries, or structured interfaces. The direction is clear. ChatGPT is evolving from “an AI that answers questions” into “an AI that creates the tools and interfaces needed to complete the work.”

2. Why are Custom GPTs moving toward Plugins?

Custom GPTs were among the most useful features of earlier ChatGPT generations. Users could create specialized AI systems by adding custom instructions, knowledge files, domain expertise, workflows, and external tools. This was powerful, but it also meant that several different functions were bundled together inside one GPT. Professional knowledge, instructions, tools, APIs, interfaces, and actions were all mixed into a single unit.

The new Plugin architecture separates those elements more clearly. When a Custom GPT is migrated, its main components can be reorganized into a Plugin structure: GPT instructions become a Plugin Skill, GPT knowledge files become Reference Files, and connected external services become Plugin Apps. This is not simply the disappearance of GPTs. It is better understood as the repackaging of specialized GPT capabilities into a larger and more modular Plugin architecture.

3. The new Plugin is very different from the old Plugin

In the earlier ChatGPT era, a Plugin was mostly a bridge to an external service. The structure was relatively simple: ChatGPT → External Service. A Plugin might call a travel service, calculator, search service, or external API. The new Plugin concept is much broader.

A modern Plugin can combine Skill, Reference Files, Apps, MCP Tools, and Extensions. A Skill contains professional instructions and workflows, while Reference Files contain detailed domain knowledge. Apps connect external services such as Gmail, Google Drive, or Calendar, and MCP Tools can connect private databases, APIs, company systems, and specialized engines. Extensions can provide custom interfaces such as forms, dashboards, file handlers, and interactive views.

This means that a Plugin is no longer merely an external connector. It can become a complete specialized AI application package.

4. What happens when a professional GPT becomes a Plugin?

In a professional GPT, the core value usually lies in its domain knowledge and reasoning rules. Consider a specialized employment and talent-matching GPT. It might contain industry and occupation ontologies, JOB, SHIFT, and TASK models, matching logic, trust rules, runtime rules, OXMV verification, and Evidence Ledger structures. When converted into a Plugin, those elements can be reorganized more systematically.

The Skill can contain integrated operating instructions, while Reference Files hold specialized knowledge. Apps can connect services such as Google Drive, Gmail, and Calendar. Custom MCP Tools can connect employment APIs, employer databases, talent databases, and mapping services. The result is more than a knowledge-based chatbot: it becomes a system combining professional intelligence, live data, and real-world action.

5. What is the difference between a migrated GPT Plugin and an external App Plugin?

This distinction is important. Some Plugins are built mainly around external applications such as Gmail, Calendar, Drive, Slack, and customer relationship management systems. Their primary value lies in accessing external data and carrying out actions. They can read, write, search, send, schedule, or update information.

By contrast, a professional GPT-derived Plugin is centered on domain knowledge and decision logic. Examples might include financial crime analysis, smart city planning, AI and semiconductor strategy, regional policy analysis, space science, and employment matching. The easiest way to understand the difference is that an external App Plugin provides the hands and feet, while a professional GPT-derived Plugin provides the specialized brain. The strongest architecture combines both.

6. How do you build a specialized Plugin?

The development process is more structured than before. The first step is to organize the domain knowledge by defining what the system must know, which ontology it uses, which rules it follows, what must be verified, and what types of decisions it can make. The second step is to create the Skill, which defines how user intent is interpreted, which knowledge files are used, how reasoning proceeds, which workflow is followed, and how output is controlled.

The third step is to add Reference Files. Detailed professional knowledge should be separated from the Skill, and these files may contain ontologies, rules, legal references, technical definitions, workflow specifications, and domain manuals. The fourth step is to add external Apps such as Google Drive, Gmail, Calendar, Slack, or enterprise tools.

The fifth step is to add MCP servers or custom tools when no existing App provides the required function. These may connect proprietary databases, government APIs, company systems, search services, or specialized engines. The sixth step is to add UI elements and Extensions, such as dashboards, candidate lists, forms, maps, file editors, and interactive result views. The development path can therefore be summarized as Knowledge → Skill → Reference Files → Apps → MCP → UI.

7. Has dot inherited the “agent role” of Custom GPTs?

Conceptually, yes—at least in part. It would be inaccurate to say that OpenAI literally split Custom GPTs into two pieces, with one half becoming Plugins and the other becoming dot. Functionally, however, that is close to what the new architecture looks like. A traditional GPT mainly behaved as “a specialist AI that responds when asked,” while dot is closer to “an AI agent that keeps working toward a goal.”

A dot can maintain context over time, use connected tools, monitor changes, and repeatedly act on the user’s behalf. For example, imagine an AI international standards Plugin containing expert knowledge. A dot could then be assigned the following task: “Check new international AI standards every week and tell me when they conflict with the standard I am developing.” In that structure, the Plugin is the specialist brain, while dot serves as an ongoing AI staff member.

8. What is Work, and how is it different from dot?

Work is also agentic, but its role is different. A useful distinction is that Work is an execution agent for completing a complex task, while dot is a persistent agent for managing an ongoing responsibility. For example, “Research the Korean construction industry over the last five years and prepare a full report” would be a suitable task for Work. By contrast, “Monitor construction policy changes every week and notify me when something important changes” would be more naturally suited to dot.

9. What is Page?

Page is one of the most important new concepts. The old Canvas experience was essentially an editable document or coding surface beside the conversation. Page extends that idea much further. A Page is a living work artifact that can continue to evolve with the user and AI.

Instead of producing a draft, a revised draft, another revision, and a final version as separate chat messages, users can continuously improve a single Page. They can ask the AI to shorten section three, add an OXMV section, leave a particular paragraph unchanged while revising the conclusion, or turn part of the document into a diagram. Page is therefore not simply another chat response. It is a persistent work product.

10. Is Space just a larger version of Canvas?

Not exactly. It is more accurate to say that Canvas evolved conceptually into Page, while Space became the larger workspace containing Pages, files, and collaborative assets. A Space can contain multiple Pages, folders, uploaded files, reference materials, team documents, and AI-generated outputs. It serves as the organizational layer above individual Pages.

For example, a single professional AI project might include a System Architecture Page, User Manual Page, OXMV Design Page, API Design Page, Release Audit Page, and Knowledge Files folder. Chat becomes the place where thinking happens. Page becomes the place where results are developed. Space becomes the place where the work is organized and preserved.

11. Why do Page and Space matter?

One long-standing weakness of conversational AI was that valuable results disappeared inside lengthy chat histories. Page and Space address that problem. They shift ChatGPT from a conversation-centric system toward a knowledge- and workspace-centric system. This is particularly important for long-term research, policy work, technical documentation, knowledge-base development, professional AI systems, and collaborative projects.

Instead of repeatedly recreating documents from chat history, users can maintain durable project artifacts. This changes ChatGPT from a temporary conversation tool into an environment where knowledge and work can accumulate over time.

12. Intelligent UI is another major change

The new generation of ChatGPT is also becoming more adaptive in how it presents results. Rather than returning only text, the system can increasingly decide whether a task would be better served by a visual comparison, chart, calculator, structured panel, diagram, or interactive control. This represents another important transition. ChatGPT is no longer only generating information; it is beginning to generate the interface required to use that information effectively.

13. Audio and meeting workflows are also expanding

Modern ChatGPT increasingly supports audio-based workflows as well. Users can bring in recorded discussions, meetings, or spoken content and turn them into transcripts, summaries, decisions, action items, and follow-up documents. Combined with meeting-oriented Plugins and Space and Page workflows, this creates a new pattern: conversation → transcription → analysis → action items → permanent project record. This is a much more operational model of AI use.

14. The new ChatGPT architecture in one view

The current structure can be understood as follows:

ChatGPT → general AI intelligence and conversation
Plugin → specialized knowledge, Skills, workflows, tools, and external Apps
dot → persistent personal AI agent
Work → execution agent for complex tasks
Page → evolving AI-assisted document
Space → workspace for Pages, files, people, and AI
Intelligent UI → adaptive interface that creates the right view or tool for the task

This division of roles is the key to understanding the new ChatGPT ecosystem.

15. What has really changed?

The biggest change is not simply a more powerful model. The deeper change is the transition from Question → Answer to Goal → Knowledge → Tools → Agent → Execution → Artifact → Ongoing Management. This represents a fundamentally different way of thinking about AI. AI no longer exists only to explain something.

AI can perform work, use tools, coordinate Apps, create artifacts, monitor ongoing tasks, maintain project context, and support repeated decision-making. Its role is expanding from answering individual questions to participating in a continuing workflow.

ChatGPT is becoming a working environment, not just a chatbot

Recent changes may feel confusing because the ChatGPT we once knew is being reorganized. Professional GPT capabilities are moving toward Plugins, while domain knowledge is being separated into Skills and Reference Files. External services are being connected through Apps and MCP. Persistent responsibilities are increasingly suited to dot, while complex one-time tasks are increasingly suited to Work.

Long-lived results are being maintained in Page, while multiple Pages and files can be organized in Space. The result is that ChatGPT is moving toward something much larger than a chatbot. It is becoming an AI brain, specialized knowledge, agents, applications, and workspaces integrated into one environment.

The question for users is therefore changing. It is no longer only, “How should I ask ChatGPT a better question?” The new question is, “Which work should be given to a Plugin, which responsibility should be assigned to dot, which task should be delegated to Work, which result should live in Page, and which project should be organized in Space?”

We are entering an era in which the real skill is no longer merely using AI. It is designing how humans and AI work together.

About the Author

Manripungchwi Jung Yeon-tae is a professional-knowledge AI planner, designer, and GPT development specialist. He currently serves as President of the Construction Industry Information Institute and develops professional GPTs across construction, law, taxation, finance, management, medicine, science, history, and philosophy. His research focuses on verified professional-knowledge AI combining ontology, XAI, the OXMV Loop, and the Evidence Ledger.

Read the Korean version

Series Overview

Medical imaging standards connect CT and MRI systems with picture archiving and communication systems and AI analysis. Hospitals must monitor model performance and access records while maintaining clinical operations during security incidents or system failures. / Korea IT Times

Medical imaging artificial intelligence can operate reliably in hospitals only when imaging equipment, picture archiving and communication systems (PACS), radiology information systems (RIS) and electronic medical records (EMR) exchange patient and examination information under consistent rules. Digital Imaging and Communications in Medicine (DICOM), the international standard for medical imaging, provides the foundation for storing and transmitting CT, MRI and X-ray images together with patient, examination and equipment data.

Transmitting images in DICOM format does not automatically ensure the accuracy or security of AI results. If hospitals use inconsistent examination names and anatomical labels, or connect images to the wrong patient record, an AI system may miss a relevant study or send it to the wrong algorithm. Without encryption, access controls, audit logs and model-version management, a standardized network can also spread errors and security threats.

Medical imaging AI performance depends on the algorithm, input data and hospital operating environment. The same product may produce different results as CT manufacturers, reconstruction methods, slice thickness, contrast use, imaging protocols and patient populations change. Hospitals need to know which images and metadata entered an AI system, which model version processed them and who received the results.

DICOM Provides a Common Language for Clinical AI

DICOM is more than a file format. It links patient identity, examination time, body part, imaging equipment and acquisition protocol to the image while defining how healthcare institutions transmit and retrieve the data. DICOMweb extends these functions to web-based environments, allowing systems to search, retrieve and store medical images.

AI-generated information, including the location and size of a lung nodule, a suspected intracranial hemorrhage and a mammography risk score, must remain accurately linked to the source image and patient examination. If results appear only in a separate application or static image file, radiologists may struggle to compare them with the original study or track the same lesion during follow-up. Structured DICOM objects allow measurements, annotations and source images to be managed within the existing reading environment.

Using the same standard does not guarantee interoperability between products. Vendors may support different DICOM functions, data fields and methods for displaying AI results. Before deployment, hospitals should review the vendor’s DICOM conformance statement and test image transmission between the product, imaging equipment and PACS.

Inconsistent examination names are another source of error. The same chest CT may be recorded under different names and codes across hospitals and scanners, making it difficult for an AI system to identify the correct study. Systems that depend on basic text-matching rules may miss an emergency examination or analyze an image outside the product’s intended use. Hospitals therefore need standardized information on the body part, imaging method, contrast use and image series, along with procedures for handling exceptions.

Hospitals must also verify how AI results appear on clinicians’ worklists. If a stroke or intracranial hemorrhage system moves a suspected case to the top of the queue but fails to deliver the alert to the responsible physician, treatment will not begin sooner. Repeated false-positive alerts can also cause alert fatigue and delay the review of other emergency cases.

Stable integration with PACS, electronic medical records and multi-institutional workflows could become a more important competitive factor than the accuracy of a single algorithm,” said Youna Kim, deputy general manager of CorelineSoft’s media team. In markets such as Germany’s lung cancer screening program, where first and second readings, prior-image comparison, follow-up and quality management are connected, companies must support the full screening process rather than a single examination.

CorelineSoft said demand is increasing for AI systems that connect multiple institutions and readers as Germany introduces low-dose CT lung cancer screening under its statutory health insurance system. However, contracted sites, completed installations, examination volumes and clinical adoption rates measure different stages of deployment. National screening programs should therefore assess AI analysis volumes, prior-image matching, interpretation times, review rates and adherence to follow-up examinations.

The effect of a system failure also increases as AI becomes more deeply embedded in hospital operations. Images must not disappear from reading worklists when a cloud connection fails or a vendor’s server becomes unavailable. Clinicians should be able to continue reading examinations without AI and identify missed studies and affected patients after service is restored.

Security Must Extend from Image Transmission to Clinical Recovery

The DICOM standard includes security mechanisms supporting encrypted communication and audit records. Transport Layer Security can protect data in transit, while audit trails can record who accessed a patient’s information, when the access occurred and what action was taken. These capabilities provide little protection if a hospital does not activate them or if legacy imaging equipment does not support the relevant security profiles.

Healthcare institutions should verify encryption, user authentication, access privileges, certificate and cryptographic-key management, and log-retention periods for each product. Contracts should define when AI vendors and cloud operators may access patient images. Remote-maintenance accounts and external network connections should remain open only for the required period and purpose, with every access recorded.

On October 8, the US Health Sector Coordinating Council Cybersecurity Working Group released its Playbook for AI Clinical Resilience. It calls for clinicians, IT and security teams, governance leaders and compliance officers to coordinate their response from the detection of an AI-related threat or failure through patient-safety review, regulatory reporting and system recovery.

The playbook treats medical imaging AI security as more than preventing data leaks. Healthcare institutions should decide in advance how clinicians will return to established workflows when AI assigns an incorrect priority or an emergency alert stops working. They also need procedures for identifying affected patients, reviewing potentially incorrect results and deciding when the system can safely resume operation. Although the playbook is not legally binding, it treats AI failure as a clinical-continuity and patient-safety issue.

Researchers have also identified a new attack method. In a study published by Scientific Reports on October 3, investigators embedded malicious instructions in the pixels of dental panoramic radiographs and tested four vision-language models: GPT-4o, Gemini 2.5 Flash, Claude Sonnet 4.5 and MedGemma 4B.

The researchers conducted 58,320 inference calls using 270 images and found that all four models were affected. Under some conditions, the attack success rate reached 62.6%. An optical character recognition-based defense that detected and removed text embedded in the image reduced the overall attack success rate from 15.6% to 0.2%.

The findings show that protecting image headers and transmission networks may not prevent commands embedded in pixels from manipulating a multimodal AI system. Hospitals using vision-language models to process images received from outside institutions or uploaded by patients should verify both the image source and any text contained within it. If the system detects an abnormal instruction, it should stop the AI analysis and refer the case for human review.

AI Performance Requires Continued Oversight After Standardization

The study does not establish that every medical imaging AI product currently used in hospitals faces the same level of risk. It was a controlled adversarial experiment using a public dental-image dataset, not an investigation of a breach at a healthcare institution. Conventional imaging AI designed primarily to detect lesions has a different attack surface from a vision-language model that interprets written instructions.

The UK government accepted all 44 recommendations of the National Commission into the Regulation of AI in Healthcare on October 6. The government plans to manage medical AI through a framework covering deployment, real-world performance, clinical responsibility, patient transparency and the full product lifecycle. It concluded that existing medical-device regulations do not adequately reflect the speed of AI development and said premarket review and postmarket monitoring should be connected.

Nabile Safdar, MD, chair of the Society for Imaging Informatics in Medicine, has described the responsible use of imaging AI as “an ongoing process rather than a single event.” He said dedicated teams must consistently apply defined procedures supported by appropriate methods and technology, identifying interoperability standards, expert participation, AI science and methodological rigor as foundations for clinical implementation.

Safdar’s assessment shows why oversight must continue after DICOM integration. Hospitals still need to manage model versions, performance changes, access logs, errors and criteria for suspending use. Standardization provides the connection to hospital systems, while institutional monitoring and governance determine whether the technology remains safe.

South Korea can use its extensive PACS infrastructure, national health insurance system and screening data to build a national performance-management framework for medical imaging AI. This would require links between product approvals and change-management information held by the Ministry of Food and Drug Safety, hospital usage records, equipment and imaging conditions, and clinical data held by the Health Insurance Review and Assessment Service. The scope of data collection and access rights must also protect patient information and commercially sensitive material.

AI procurement contracts used by hospitals and public institutions should specify more than whether a product supports DICOM. They should identify the supported data objects and examination codes, formats used to store AI results, encryption and logging requirements, model-update procedures, response times for outages and thresholds for performance deterioration. They should also define how data will be handled and how another product can replace the system when the service ends.

Standardization does not require every medical imaging AI product to operate in the same way. It creates a consistent structure linking the patient, examination, image and AI result so that hospitals can identify the cause of an error and maintain clinical records and follow-up care when a product is replaced. Medical imaging AI can function as core hospital infrastructure only when standardized connectivity, performance monitoring, cybersecurity and recovery procedures operate together.

Part 7 will compare medical imaging AI reimbursement in the United States and South Korea and examine the clinical and economic evidence required to move from regulatory approval and billing codes to actual insurance coverage.

Read the Korean version

Capital markets are changing how they assess South Korean medical artificial intelligence companies. A few years ago, overseas regulatory milestones, including U.S. Food and Drug Administration (FDA) authorizations and European CE marking, along with supply contracts with major hospitals, were considered key drivers of share prices. The ability of a technology company to enter a tightly regulated healthcare market was itself a scarce achievement. Increasingly, valuation depends on how extensively authorized products are used in clinical practice and how that use translates into revenue.

For remote patient monitoring, key indicators include installed beds and penetration within hospitals; for medical imaging AI, they include the number of examinations analyzed and ongoing usage volumes. Overseas operations also require scrutiny of reimbursement arrangements, claims submission, contract renewal rates and the timing of revenue recognition. The industry has reached a stage where technically validated products are judged on their ability to generate sustainable revenue in clinical practice. In a recent analysis of Seers Technology, IBK Securities highlighted continued expansion of domestic monitoring beds and the timing of revenue recognition from its United Arab Emirates (UAE) business as key variables. Because the scope of a supply contract can differ from actual operations, investors need to distinguish the proportion of installed beds covered by paid services and the duration of use at each hospital.

Coreline Soft illustrates this shift in medical imaging AI. The company said it secured 77 contracts with new customers in Germany through July this year. That represents a substantial increase from approximately 10 during the whole of last year. As low-dose computed tomography (CT) lung cancer screening for high-risk groups entered Germany’s statutory health insurance system, participating healthcare providers also expanded their screening infrastructure.

Germany’s lung cancer screening program requires more than AI that identifies an individual pulmonary nodule. It must connect multiple institutions and readers across initial and second readings, comparisons with previous images, follow-up and quality assurance. A market is taking shape in which medical imaging AI operates as part of the screening workflow. To support repeat screening, products must integrate reliably with hospitals’ picture archiving and communication systems (PACS) and reading workflows, enabling consistent comparisons between previous and current examinations.

Conversion to Clinical Use and Cash Collection Matters More Than Contracts Alone

Inclusion in screening programs also affects how medical AI companies earn revenue. While many earlier contracts involved a one-time sale of software licenses, payment models may develop around screening volumes or contracts renewed after a defined period of use. In a September 23 site-visit report on the outlook for recurring revenue, SK Securities highlighted the possibility that Coreline Soft’s revenue mix would shift from one-time sales toward recurring income. It projected that installations this year would lead to per-examination revenue next year. This remains a brokerage forecast rather than a confirmed financial result.

In the United States, indicators of existing customer usage and subscription longevity are emerging alongside new regulatory authorizations. Coreline Soft said its subscription relationship with medical imaging post-processing specialist 3DR Labs is in its fourth year, with annual usage increasing by 60%. This indicates how extensively an existing customer continues to use the product. The company also reported that overseas revenue rose 123% year on year in the first half, accounting for 61.9% of total revenue, while its operating loss narrowed by 21%. These provide early indicators for assessing whether its overseas installed base is beginning to translate into revenue.

Contract counts, installed sites, active sites and AI analysis volumes measure different things. Overseas revenue should also be separated into software license sales, per-examination charges and maintenance revenue. Commercialization can be examined across regulatory authorization, contracting, installation, actual use, reimbursement and billing, contract renewal, and additional product deployment. Authorization, contracts and installation establish the foundation for market entry, while usage, billing, renewals and further deployments indicate whether a product has become embedded in hospital operations and generates sustainable revenue.

Differences in company value emerge from both performance at each stage and conversion to the next. Investors need to examine the proportion of authorized products that secure contracts, contracted institutions that complete installation and begin operations, and AI analyses that translate into insurance claims and revenue. Usage at each institution and contract renewal rates provide a basis for assessing the durability of recurring revenue. Adoption of additional products indicates the scope for expanding business within an existing customer base.

Contract value does not automatically become revenue or cash. Hospital installation schedules, acceptance procedures, insurance claims and payment processes at governments and public institutions can delay revenue recognition and collection. For overseas contracts, actual use may begin only after local regulatory requirements, system integration and staff training are completed, extending the interval between an announcement and revenue generation. Contract size therefore needs to be considered alongside the time required to begin operations and the collection period after revenue is recognized.

The scale of recurring revenue must also be distinguished from its profitability. Higher analysis volumes can bring higher costs for cloud infrastructure, computing, hospital-specific integration, security management, clinical staff training and local customer support. Revenue per active institution and earnings per examination must be assessed alongside these operating costs to judge the quality of recurring revenue. This framework applies to Coreline Soft, remote patient monitoring providers including Seers Technology, and medical imaging AI companies in Korea and overseas. Comparable assessments of commercialization require companies to report contracted beds and hospitals separately from actual usage, insurance claims and cash collection.

Reimbursement and Hospital Integration Determine Recurring Revenue

Medical AI competitiveness depends on algorithmic accuracy, reliable use within hospital workflows and connection to established payment systems. Continued use and contract renewal by existing customers shape long-term revenue. In the United States, sustainable payment arrangements and integration into clinical practice are becoming concrete challenges for wider adoption. In July, the Journal of the American Medical Informatics Association (JAMIA) published “Translating Innovation into Payment: A.I. Reimbursement in U.S. Health Care,” which analyzed how payment for clinical AI is distributed across existing fee-for-service arrangements, additional payments and procedure codes. The analysis argued that current payment mechanisms do not adequately reflect the value of workflow integration and repeated use.

The U.S. Centers for Medicare & Medicaid Services (CMS) uses Current Procedural Terminology (CPT) and Healthcare Common Procedure Coding System (HCPCS) codes, along with New Technology Add-on Payments (NTAP), in payment arrangements for certain AI-enabled medical services. The existence of a billing code, however, does not automatically guarantee reimbursement. Actual payment depends on meeting requirements for patient indications, medical necessity, clinician qualifications, payer coverage policies and billing conditions. Assessments of a company’s prospects must therefore distinguish the availability of a code from whether an insurer pays the claim.

Insurance reimbursement and inclusion in national screening programs create a foundation for recurring use while increasing dependence on policy. Changes in covered populations, payment rates or provider eligibility can alter expected examination volumes and revenue. Growth opportunities from entry into established healthcare systems should be assessed alongside a company’s pricing power and the duration of reimbursement coverage. Revenue concentration in a particular country or program is another measure of long-term stability.

In Europe, clinical adoption and hospital system integration have been identified as major priorities for medical AI. According to “Artificial Intelligence Is Reshaping Health Systems: State of Readiness Across the European Union,” published in April by the World Health Organization’s Regional Office for Europe (WHO/Europe), 74% of the European Union’s 27 member states use AI diagnostic tools, including those for medical imaging and disease detection. WHO identified workforce training, data governance and common standards for safe deployment as key priorities. Clinical use requires an operational framework in which healthcare professionals understand the technology’s limitations and potential errors and remain accountable for final decisions.

The European Health Data Space (EHDS) includes medical images and related reports among the priority categories of health data for exchange. Interoperability and access-logging requirements for electronic health record systems are scheduled to take effect in phases from 2029 and 2031, depending on the data category. EHDS establishes a framework for secure data exchange between healthcare institutions and countries, distinct from defining medical imaging technical standards themselves. Data portability when institutions replace systems or add another company’s AI can reduce supplier dependence and duplicative integration costs.

Deep integration with hospital systems can strengthen contract retention while creating a risk of vendor lock-in. The ability to transfer data and previous analysis results to another product is important. So is the ability to add external algorithms. Long-term competitiveness should be assessed through a company’s ability to secure renewals on the strength of standardized integration and demonstrated clinical outcomes.

Coreline Soft is a representative example of this shift in medical imaging AI. Its expanding German contracts, U.S. subscription relationship and overseas revenue growth indicate that the sector is moving beyond regulatory authorization and initial deployment toward sustained use. The next stage of assessment is to examine the proportion of contracted institutions actively using the product and how examination volumes translate into insurance claims, revenue recognition, contract renewal and cash collection. Assessing recurring revenue profitability also requires accounting for cloud infrastructure, system integration, security, clinical staff training and customer support costs. Alongside contract scale, the long-term value of medical AI companies depends on whether products embedded in clinical practice can generate sustained cash flow and improve patient outcomes.

Read the Korean version

Re began selling its products at 50 Costco stores across the United States on October 7, marking its first major expansion into Costco’s offline retail network. The move follows sales through Costco’s online store, Amazon and TikTok Shop.

The brand has already seen strong demand through Costco’s online platform. Its Cell to Cell Cream sold out ahead of schedule in March, followed by the Cell to Cell Essence in May. The early sellouts prompted Re to broaden its presence into physical stores and increase direct access to local consumers.

Alongside its Costco expansion, Re has introduced its new Celestial Collagen Cream to the U.S. market. The product uses collagen water instead of purified water as its base and contains eight types of collagen and 10 types of peptides. The company said its proprietary Re Celestial ingredient was formulated to help improve skin density and elasticity.

The expansion comes as Hil​​erby seeks to diversify its sales channels and strengthen its overseas business. In Korea, the company has expanded Re’s offline presence through The Hyundai Seoul’s The Wellness House and Musinsa Beauty Hongdae this year.

Hilerby is also looking beyond the United States, with plans to expand into Canada and other North American markets as well as Europe. The company is positioning Re as a key brand for building its presence in the global K-beauty market.

The overseas expansion is also important for Hil​​erby as it works to improve profitability. According to previous reports citing regulatory filings, Hil​​erby recorded 8.5 billion won in revenue in the first half of this year, down 25.4% from 11.4 billion won a year earlier. Its net loss narrowed to 2.98 billion won from 3.53 billion won, but the company remained in the red.

The expansion of Re’s U.S. offline distribution network is therefore expected to be an important test of whether its online momentum can translate into higher sales and improved earnings.

A Re representative said the Costco expansion will provide an opportunity to showcase the brand’s collagen skincare technology directly to U.S. consumers, adding that the company plans to continue introducing differentiated K-beauty products and expanding its global distribution network.

Hyundai E&C said Thursday that it has signed a contract with Westinghouse to participate in the first phase of the front-end engineering design, or FEED 1, for the Dutch government’s planned construction of two large-scale nuclear reactors.

The project is being developed under the Netherlands’ renewed push to expand nuclear power. Hyundai E&C will work alongside Westinghouse on the proposed deployment of two AP1000 reactors, focusing on the early engineering and project planning needed to move the development toward the construction stage.

Rather than limiting its role to construction, Hyundai E&C will be involved from the early stages of the project. Its responsibilities include assessing potential sites, developing an overall master plan and establishing a strategy for project execution.

The two companies will also examine Dutch regulatory requirements, licensing procedures and technical standards that could affect the design and construction of the reactors. Risk management measures and localization strategies will also be considered as part of the FEED work.

The early involvement is strategically important as the Netherlands evaluates potential sites and prepares the framework for its new nuclear program. Westinghouse has emerged as one of the technology candidates for the project, and the FEED phase is expected to help translate the proposed AP1000 design into a project plan suited to local site and regulatory conditions.

Hyundai E&C has also been working to establish a local industrial network ahead of the project. In June, the company and Westinghouse jointly hosted a supplier symposium in the Netherlands, bringing together government representatives, industry stakeholders and local companies to discuss potential cooperation and the development of a local supply chain.

The company sees the Dutch project as part of a broader push to expand its nuclear business across Europe. Hyundai E&C has been strengthening its presence in the region through various nuclear-related initiatives, including projects and preliminary studies connected to markets such as Bulgaria, Slovenia and Finland.

The FEED contract could give Hyundai E&C an opportunity to build an early understanding of local regulations, site conditions and construction requirements, potentially strengthening its position for subsequent stages of the Dutch nuclear program.

The company is also expanding its capabilities beyond conventional nuclear plant construction. Its nuclear strategy covers the wider lifecycle of nuclear facilities, including decommissioning and spent-fuel management.

A Hyundai E&C official said the company plans to leverage its experience in nuclear engineering and construction to support the successful development of the Dutch project while continuing to pursue new opportunities across Europe’s expanding nuclear market.

The Americas Cup will take place on October 11 at Chrysalis Hall at Toronto Metropolitan University, with matches scheduled to begin at 3 a.m. KST. The event marks the first time a SWC regional cup has been held in Canada, adding a new destination to the tournament’s international circuit.

Eight players have secured their places in the Toronto showdown after making it through the Americas regional qualifiers. The quarterfinal bracket features SAPPYYY against TRUEWHALE, JIWOOFAN against ZEDORUS, COISO against ZEZAS, and THOMPSIN against MONAGUA.

Several players arrive in Toronto with storylines of their own. TRUEWHALE enters the tournament as a two-time defending regional champion and will be looking to extend that run once again. ZEZAS, meanwhile, has established himself as one of South America’s rising forces after making the transition from a Brazilian newcomer to a leading contender in the region.

THOMPSIN is another player attracting attention after returning to the main stage following an extended absence. With established champions and emerging challengers sharing the bracket, the Americas Cup is set to offer a competitive race for the three World Finals berths.

The tournament will use a best-of-five single-elimination format. Following the quarterfinals and semifinals, the three highest-placing players will earn tickets to the SWC2026 World Finals.

Fans around the world will be able to follow every match through the official Summoners War Esports YouTube channel, with multilingual commentary available throughout the broadcast.

The Americas Cup is part of a wider regional championship schedule leading into the season finale. Com2uS will hold the Asia-Pacific Cup in Bangkok on October 17, following the European Cup, which kicked off the regional finals on September 26.

The 12 players who emerge from the regional cups will then meet at the SWC2026 World Finals on November 14 at KBS Arena in Seoul. They will compete for the title of the 10th global champion in the history of the tournament.

First launched in 2017, SWC has grown into the premier international esports competition for ‘Summoners War: Sky Arena,’ Com2uS’ globally popular monster-collection strategy game. Players build teams from a wide range of monsters and compete through tactical turn-based battles.

As SWC enters its 10th edition, the 2026 season is bringing the competition to new locations while continuing its search for the best ‘Summoners War’ player in the world.

The SWC2026 Americas Cup will be streamed live on the official Summoners War Esports YouTube channel. Additional tournament information is available through the official Summoners War community channels.

Zurich-based Enso said on October 7, 2026, that xStocks are available on Arbitrum through its API and real-world asset app. Wallets, fintech companies and trading platforms that already use Enso’s API can add access to the tokenized equities without building separate routing infrastructure, the company said. Eligible users can also trade xStocks through the Enso RWA App.

Enso provides infrastructure for transactions across blockchain networks, including swaps and decentralised finance activity. The company said the xStocks integration extends those tools to tokenized equities on Arbitrum. Applications using its API can offer access through their existing connection, while the RWA App provides a direct route for eligible users.

“Tokenization solved the supply side of real-world assets, but what’s been missing is distribution,” said Connor Howe, Enso’s co-founder and chief executive. He said wallets and other applications need liquidity, routing and execution tools to make tokenized assets available where users already transact. Enso also cited its work on tokenized stock and metals execution for ether.fi Cash, a minting role for Ondo Finance, and infrastructure used by Anchorage Digital’s Porto institutional terminal.

Peter Haymond, head of DeFi partnerships at Arbitrum, said shared execution infrastructure could help developers add tokenized equities as more real-world assets become available on the network. Access to xStocks through the Enso RWA App is available now, according to Enso. Developers can add the assets on Arbitrum through the Enso API.

The South Korean technology giant said Wednesday that it expects consolidated sales of 195 trillion won and operating profit of 107.4 trillion won for the third quarter of 2026. Sales surged 126.6% from a year earlier, while operating profit jumped 782.5%, with both figures marking quarterly records.

Operating profit also exceeded market expectations of around 106 trillion won and surpassed the previous quarterly record of 89.49 trillion won set in the second quarter. Samsung has now broken its quarterly operating profit record for four consecutive quarters since the fourth quarter of last year.

The scale of the latest earnings is unprecedented among global technology companies. Samsung's quarterly operating profit is significantly higher than that of Nvidia, the world's largest company by market capitalization, which reported operating profit equivalent to roughly 85 trillion won in its most recent quarter.

The extraordinary performance was driven primarily by Samsung's semiconductor business, which has benefited from a sharp increase in demand for memory chips used in AI data centers.

Although Samsung has yet to disclose detailed results by business division, analysts estimate that the Device Solutions division generated more than 100 trillion won in operating profit during the quarter. The memory business is believed to have accounted for the overwhelming majority of those earnings.

The memory market has entered a powerful upswing as global technology companies rapidly expand AI infrastructure. High-bandwidth memory, or HBM, has emerged as a critical component for AI accelerators, while demand for conventional DRAM has also strengthened as data-center operators increase server capacity.

At the same time, memory manufacturers have been shifting production capacity toward higher-margin HBM products. This has tightened supplies of conventional DRAM and contributed to broad-based price increases across the memory market.

Industry data indicates that conventional DRAM contract prices rose by double digits during the third quarter. Although the pace of increase slowed from the previous quarter, prices continued to rise as supply remained constrained.

Samsung has benefited from both higher shipment volumes and rising chip prices. The combination has sharply improved the profitability of its memory operations, with the business believed to have achieved margins rarely seen in the semiconductor industry.

HBM4 also became an increasingly important contributor to Samsung's performance during the quarter. The company has begun ramping up production of its next-generation HBM as demand from AI chipmakers accelerates.

The launch of Nvidia's next-generation AI platform has further increased demand for HBM4. Samsung is supplying HBM4 for the new generation of AI systems, giving the company an opportunity to strengthen its position in the rapidly expanding AI memory market.

HBM4 could also have implications beyond Samsung's memory business. The company is using its own advanced foundry technology to manufacture base dies for HBM4, creating potential synergies between its memory and contract-chip manufacturing operations.

As HBM production expands, demand for related logic dies, advanced packaging and foundry services could increase as well. That could provide a boost to Samsung's foundry business, which has struggled in recent years to match the profitability and growth of its memory operations.

The semiconductor boom, however, has created a mixed picture across Samsung's broader business portfolio.

Samsung's Device Experience division, which includes smartphones, televisions and home appliances, has faced rising component costs as memory prices have surged. Higher memory prices benefit the semiconductor division but increase production costs for businesses that use memory chips in finished products.

The pressure has been particularly significant in smartphones and consumer electronics, where fierce competition makes it difficult for manufacturers to pass higher component costs directly on to customers.

As a result, analysts expect the DX division to have remained under significant pressure during the third quarter, following its first quarterly operating loss earlier this year.

The contrast highlights the unusual nature of Samsung's latest earnings. The company did not achieve its record profit through uniform growth across all businesses. Rather, the semiconductor division generated an extraordinary amount of profit that more than offset weakness elsewhere.

The outlook for Samsung's semiconductor business remains strong, at least for the near term.

AI data-center investment continues to accelerate globally, while semiconductor manufacturers face limitations in expanding production capacity quickly. Building new fabrication facilities requires substantial capital and years of preparation, meaning supply is unlikely to catch up with demand immediately.

This imbalance has encouraged major customers to secure memory supplies through long-term agreements. Technology companies are increasingly seeking guaranteed access to memory chips in anticipation of continued shortages, strengthening the bargaining position of major suppliers.

The trend could help extend the current memory upcycle well into next year.

Samsung's accumulated earnings have also put the company on track for another major annual milestone. With operating profit already exceeding 250 trillion won through the first three quarters, the company could potentially surpass 300 trillion won in annual operating profit if the current semiconductor boom continues.

Still, maintaining such extraordinary profitability will not be easy.

The memory industry remains cyclical, and an eventual increase in production capacity could put downward pressure on chip prices. Samsung must also maintain its technological lead in HBM while competing aggressively with other major memory manufacturers for AI customers.

The company faces challenges beyond memory as well. Improving the profitability of its foundry and system semiconductor businesses remains a key priority, while its smartphone, television and home appliance operations need to recover margins as component costs rise.

For now, however, Samsung's third-quarter results mark a historic moment for the company and the global technology industry.

A quarterly operating profit exceeding 100 trillion won demonstrates how dramatically the economics of the semiconductor industry have changed in the age of AI.

The question for investors is no longer whether Samsung can reach the 100-trillion-won quarterly profit threshold. It is whether the company can sustain that level of profitability as AI infrastructure investment continues to reshape the global semiconductor market.

Series Overview

Episode 1 | Why the World Took Notice of Korea's Seowon

Episode 2 | An Hyang and Sosu Seowon — The Beginning of Korean Education

Episode 3 | Yi Hwang and Dosan Seowon — How Learning Shapes Character

Episode 4 | Ryu Seong-ryong and Byeongsan Seowon — Leadership That Protected a Nation

Episode 5 | Yi Eon-jeok and Oksan Seowon — Philosophy Perfected in Nature

Episode 6 | Kim Goeng-pil and Dodong Seowon — The Roots of Moral Scholarship

Episode 7 | Jeong Yeo-chang and Namgye Seowon — The Seonbi Spirit Beyond the Capital

Episode 8 | Kim In-hu and Pilam Seowon — The Enduring Power of Integrity

Episode 9 | Choe Chi-won and Museong Seowon — A Philosophy of Harmony

Episode 10 | Kim Jang-saeng and Donam Seowon — Ritual, Community, and Everyday Life

Episode 11 | Are Seowon Relics of the Past—or Schools for the Future?

The series concludes by revisiting Lee Bae-yong's efforts to achieve UNESCO World Heritage inscription while reflecting on what the Seonbi tradition can offer education, leadership, and humanity in the age of artificial intelligence.

By Jung Yeon-Tae
President, Korea Information Service of the Construction Industry

1. A Scholar’s Heart Written Upon the Brush Rock

Piram Seowon stands in Jangseong, South Jeolla Province, near open fields and settled communities. Unlike academies hidden deep within dramatic mountain valleys, it remains closely connected to the everyday landscape of village life. Its broad courtyard, orderly buildings, low walls, and open sky suggest that learning was not intended to withdraw permanently from society.

Piram Seowon was established to honor the learning and moral integrity of Haseo Kim In-hu, one of the leading Neo-Confucian scholars of mid-Joseon Korea. The academy was first founded in 1590 at Gisan-ri in Jangseong, thirty years after Kim’s death. It was destroyed during the Japanese invasion of 1597, rebuilt in 1624, granted the royal name “Piram Seowon” in 1662, and moved to its present location in 1672.

The name “Piram” may be understood through the characters for “brush” and “rock.” The brush symbolizes learning, writing, and cultivated expression. The rock suggests firmness, integrity, and an unshakable moral foundation. A brush may move softly across paper, but the principle written by that brush must remain firm like stone.

The name of the academy therefore reflects the life of Kim In-hu himself. He was a scholar, poet, teacher, and government official. Yet above all, he was a man who refused to use learning merely for advancement. He sought to embody it in loyalty, self-restraint, teaching, and moral choice. Piram Seowon preserves that spirit through architecture, ritual, documents, and education.

2. Haseo Kim In-hu, the Great Scholar of Honam

Kim In-hu was born in Jangseong in 1510 and died in 1560. His courtesy name was Huji, and his pen names included Haseo and Damjae. He was later given the posthumous title Munjeong and became one of the eighteen Korean Confucian sages enshrined in the national Confucian shrine.

His intellectual reputation rested upon three closely connected qualities: Neo-Confucian learning, moral integrity, and literary achievement. For Kim, these were not separate accomplishments. His writing emerged from learning. His learning was tested through moral conduct.

His integrity was revealed through the choices he made in political life. He did not believe that scholarship existed only in books or examinations. It had to govern the relationships between ruler and subject, parent and child, teacher and student, and friend and friend. Learning was meaningful only when it shaped human conduct.

For this reason, Piram Seowon is more than a memorial building. It preserves the life of a scholar in whom knowledge, character, and action were understood as one continuous path.

3. The Intellectual Lineage of the Sarim

Kim In-hu studied under the scholar Kim An-guk, who belonged to the intellectual tradition that developed from Kim Jong-jik and the early Sarim movement. Through teachers, disciples, lectures, letters, and regional academies, Neo-Confucian learning spread from one generation to another. It was not created by isolated thinkers. It grew through communities of inquiry in which scholars questioned texts, debated principles, corrected one another, and observed the lives of respected teachers.

Kim In-hu inherited this tradition and helped establish it firmly in the Honam region. He did not study the classics merely to memorize established interpretations. He sought to examine whether the teachings of the sages were present in his own mind and conduct. In Neo-Confucian thought, study meant both gaining understanding and removing selfish desire.

The learner did not become morally greater by accumulating information alone. He became more fully human by reducing arrogance, prejudice, greed, and self-deception. The buildings of Piram Seowon were later arranged to support precisely this form of education.

4. Teacher of the Crown Prince

Kim In-hu passed the civil service examination in 1540 and entered government. He served in offices including the Office of Special Advisers and the Crown Prince Tutorial Office. In the latter position, he taught the future King Injong.

A tutor to the crown prince did more than explain classical texts. He was responsible for shaping the moral understanding of a future ruler. He taught how power should be exercised, how a ruler should listen to criticism, how he should treat his people, and how personal desire should be restrained for the sake of public responsibility.

The relationship between Kim In-hu and Injong therefore became more than an ordinary relationship between subject and prince. Injong respected Kim’s learning and character, while Kim hoped that the prince would become a virtuous ruler. A surviving tradition associated with Piram Seowon tells that Injong, while still crown prince, gave Kim In-hu a painting of bamboo that he had personally created. Materials related to this ink-bamboo image have long been preserved at the academy.

Bamboo was a powerful symbol in Confucian culture. Its hollow interior represented humility. Its upright stem represented integrity. Its evergreen color suggested constancy. The painting can therefore be understood as an expression of trust between a student and his teacher.

5. The Brief Reign of King Injong

King Injong ascended the throne in 1544 following the death of King Jungjong. He hoped to correct political abuses and restore the position of Sarim scholars, but his reign lasted only about eight months. He died in 1545. For Kim In-hu, Injong’s death was not simply the death of a monarch.

He had lost a former student and a ruler in whom he had placed hopes for moral government. After Injong’s death, the throne passed to the young King Myeongjong, and factional conflict among royal relatives led to the Eulsa Literati Purge. Kim withdrew from political life and returned to Jangseong.

He devoted himself to scholarship, poetry, self-cultivation, and the education of disciples. Although government positions were later offered to him, he did not fully return to public office. His withdrawal was not simply an escape from responsibility. It was a moral judgment about the conditions under which service was possible.

Korean historical accounts remember Kim especially for his loyalty to Injong. Each year, on the anniversary of the king’s death, he mourned the ruler he had taught. A famous poetic image describes the sound of mourning filling the mountains every seventh month. His loyalty was not directed toward the political power of the living. He remained distant from those who currently controlled the court, while preserving loyalty to the king and principles in which he believed.

6. Withdrawal in an Age of Political Purge

The Eulsa Literati Purge of 1545 occurred during intense conflict among royal in-law families after the accession of the young King Myeongjong. Many officials and Sarim scholars were executed, exiled, or removed. For a Confucian scholar, the decision to enter or leave public office was a central moral question.

To enter government was to attempt to practice the Way in public life. To withdraw was to preserve principle when the political moment no longer allowed righteous service. The decision was not supposed to depend on personal advancement. It depended on whether moral action remained possible.

Kim In-hu’s life became an example of principled withdrawal. He left the center of power, but he did not abandon responsibility. He taught students, cultivated scholarship, and helped shape the intellectual culture of the Honam region. He withdrew from the political center without withdrawing from the moral center.

Piram Seowon therefore challenges the idea that remaining within an institution is always the most responsible choice. At times, withdrawal may be the only way to preserve integrity and create another form of public value.

7. Heaven’s Mandate and the Human Mind

Kim In-hu’s Neo-Confucian thought attempted to understand the universe and the human person within a single moral order. His intellectual work on the Diagram of Heaven’s Mandate explored the relationship among cosmic principle, human nature, and moral responsibility. In Neo-Confucianism, Heaven’s Mandate did not refer only to fate imposed from outside. It referred to the moral principle present within human nature and all things.

When the mind was not obscured by selfish desire, its original moral clarity could become visible. The purpose of learning was therefore not endless accumulation of external information. It was the careful examination of the mind, the removal of prejudice and desire, and the restoration of one’s proper human nature. For Kim In-hu, cosmology was inseparable from ethics.

A person could not claim to understand the principles of the universe while failing to practice loyalty, trust, reverence, and responsibility in ordinary relationships. Education at Piram Seowon reflected this understanding. Students read texts, but they also observed their speech, conduct, habits, and relationships. Knowledge became real only when embodied in life.

8. The Formation of Honam Neo-Confucian Culture

Kim In-hu played a major role in the growth of Neo-Confucian learning in the Honam region. His students and associates worked throughout Jangseong, Gwangju, Damyang, and neighboring areas. They built communities of study through academies, pavilions, correspondence, poetry, and philosophical debate.

Honam Neo-Confucianism was not merely a provincial copy of scholarship produced in the capital. It developed through the region’s own landscapes, families, educational networks, and political experiences. Kim In-hu became one of its central moral and intellectual figures. Later generations praised him for excellence in learning, integrity, and literature.

Piram Seowon enshrines Kim as its principal figure and also honors his disciple and son-in-law, Yang Ja-jing. Official sources describe Kim as a leading Confucian master of Honam and identify Yang as the scholar enshrined alongside him. Yang Ja-jing was added to the shrine in 1786. His inclusion transformed the academy into a visible representation of intellectual succession from teacher to disciple.

A tradition survives only when later generations do more than repeat the words of a teacher. They must reinterpret and embody those teachings in their own time.

9. Destruction, Reconstruction, and Relocation

Piram Seowon was first founded in 1590 at Gisan-ri. Only seven years later, it was destroyed during the Japanese invasion of 1597. War destroyed not only people and settlements. It burned books, printing blocks, schools, shrines, and the institutional memory of communities.

The scholars of Honam refused to abandon Kim In-hu’s legacy. The academy was rebuilt in 1624 at Jeungsan-dong and moved once more in 1672 to its present site because of concerns about flooding and location. During this relocation, an old government guesthouse from Jinwon was moved and reused as Cheongjeoldang, the main lecture hall, while Hwagyeonnu Pavilion was constructed at the entrance.

The history of Piram Seowon is therefore a history of movement and restoration. The site changed, but the meaning did not. When buildings burned, they were rebuilt. When the location proved unstable, the academy was moved. Cultural heritage survives not because it remains untouched, but because communities repeatedly choose to protect and renew it.

10. Hwagyeonnu Pavilion and the Open, Impartial Mind

The first major structure encountered at Piram Seowon is Hwagyeonnu Pavilion. Its lower level serves as a gate, while the upper level provided a place for rest, discussion, and social exchange. The name “Hwagyeon” derives from a Neo-Confucian expression meaning to open the mind widely and become great, public-minded, and impartial.

Passing beneath the pavilion symbolized more than entering an enclosed compound. The visitor was expected to leave behind the noise and selfish concerns of ordinary life and enter a space of learning and reflection. The pavilion’s signboard is traditionally attributed to the eminent scholar Song Si-yeol, who greatly respected Kim In-hu’s moral legacy.

The upper pavilion was also a place of leisure. Scholars met guests, composed poetry, viewed the landscape, and restored their minds. Confucian cultivation did not consist only of discipline and restraint. It also required moments of openness in which nature, friendship, and cultivated enjoyment allowed the mind to recover balance.

11. Cheongjeoldang, the Hall of Pure Integrity

Beyond the pavilion stands Cheongjeoldang, the central lecture hall. Its name means “Hall of Pure Integrity.” It recalls Kim In-hu’s decision to preserve loyalty and moral principle rather than pursue political advantage.

Students gathered here to read the classics, debate ideas, attend lectures, and conduct the academy’s major meetings. The building had originally served as a government guesthouse in Jinwon before being moved and converted into the academy’s lecture hall. Cheongjeoldang has an unusual orientation.

Rather than facing only toward the entrance pavilion, it is directed toward Udongsa, the shrine behind it. The student dormitories were also arranged so that the shrine remained visually and symbolically present in the daily life of the academy. This was architecture as moral pedagogy. The purpose of learning was to approach the example of the teacher.

The books read in the lecture hall had to lead toward the integrity represented in the shrine. Cheongjeoldang therefore expressed a simple principle: Scholarship was not the skill of producing elegant words. It was the formation of a clear mind and upright character.

12. Jindeokjae and Sung-uijae: Life as Education

On either side of the lecture hall stand the eastern and western dormitories, Jindeokjae and Sung-uijae. The name Jindeokjae refers to advancing in virtue. Sung-uijae refers to honoring righteousness. The buildings’ names reveal the purpose of seowon education.

Students slept, read, ate, and lived together in these spaces. Class time and personal life were not sharply separated. How a student rose in the morning, treated elders, spoke to companions, conserved shared resources, and controlled anger were all considered part of learning.

Neo-Confucian ethics understood character as the result of repeated habits. Great virtue did not suddenly appear in dramatic moments. It was formed through small acts of reverence, restraint, honesty, and consideration. The dormitories therefore taught students how to live with others and govern themselves.

The seowon was not only a place for transmitting knowledge. It was a place for forming human beings.

13. Udongsa Shrine and Gyeongjanggak Repository

At the rear of the academy stands Udongsa, the shrine where Kim In-hu is principally enshrined and Yang Ja-jing is honored alongside him. The complex follows the general Korean seowon principle of placing educational spaces in front and the shrine behind. Yet Piram Seowon places especially strong visual emphasis on the shrine, as the lecture hall and dormitories remain oriented toward it.

Memorial ritual was not simply worship of the dead. It was a form of education. Through ritual, later scholars remembered the life and learning of the honored teacher and examined their own conduct.

Near the shrine stands Gyeongjanggak, the repository for precious books, documents, and cultural materials. Piram Seowon preserves woodblocks of Kim In-hu’s writings, materials related to the ink-bamboo painting, and records concerning the operation of the academy.

These materials reveal that the seowon was not an abstract philosophical institution. It was an active community involving teachers, students, ritual, finance, labor, property, books, and administration.

14. Documentary Heritage and the Daily Life of a Seowon

One of Piram Seowon’s most significant features is its extensive documentary heritage. The collected documents of the academy comprise fifteen volumes and sixty-five sheets and have been designated as a Korean Treasure. The collection includes records of academy directors, student rosters, property administration, institutional rules, and the management of enslaved people.

These documents allow researchers to reconstruct the actual operation of a Joseon academy. A seowon needed to administer land, select students, appoint officials, maintain books, prepare ritual materials, and manage finances. It functioned as a small educational institution, ritual center, and local administrative community.

Records preserved at Piram Seowon include the register of directors, student records, property documents, and the Nobibo, which organized enslaved persons in a genealogical format. From a modern perspective, the records of slavery are also evidence of the hierarchy and inequality of Joseon society. Cultural heritage should not be interpreted only through beautiful buildings and noble ideals.

The social order, privileges, and labor relations that supported institutions must also be examined. The true value of documentary heritage lies not in idealizing the past. It lies in revealing both its achievements and its injustice, allowing the present to decide what should be preserved and what must be overcome.

15. The Ink Bamboo of Injong

Among the materials associated with Piram Seowon, the story of the ink-bamboo painting is especially moving. According to tradition, Crown Prince Injong gave Kim In-hu a bamboo painting that he had drawn himself. Kim carefully preserved it, and later generations transmitted its meaning through carved woodblocks and related records.

Bamboo held deep meaning in East Asian scholarly culture. Its stem is upright. Its interior is empty. It remains green through the seasons and bends without easily breaking. It symbolized integrity, humility, constancy, and resilience.

The gift therefore reveals the relationship between a student and his teacher. The prince recognized the teacher’s upright character in the image of bamboo. The teacher, in turn, never forgot the trust and hopes of his student. Kim’s mourning after Injong’s early death was not merely a political expression of loyalty.

It also contained the grief of a teacher who had lost a student and the sorrow of a scholar whose hope for moral government had disappeared. At Piram Seowon, the bamboo image becomes more than art. It is a record of trust connecting teacher and student, politics and learning, loyalty and grief.

16. Survival During the Abolition of Seowon

By the late Joseon period, the number of seowon had expanded rapidly. Some continued to provide education and ritual, while others became centers of local privilege, factional authority, and tax exemption. In the late nineteenth century, Heungseon Daewongun ordered the closure of most academies. Only forty-seven were allowed to remain.

Piram Seowon was among the survivors. Official tourism sources identify it as an academy that remained because of its long history and its association with Kim In-hu’s learning, integrity, and literary achievement. Its survival reflected Kim’s high status, his later enshrinement in the national Confucian shrine, and Piram Seowon’s importance within Honam.

Yet survival alone does not justify every aspect of the institution. Seowon were centers of moral education, but they also existed within a hierarchical social system. A mature understanding of heritage must hold both truths together. The moral integrity represented by Kim In-hu deserves preservation.

The privileges and inequalities of feudal society require critical reflection. Cultural inheritance does not mean recreating the past exactly as it was. It means connecting its finest principles to the universal values of the present.

17. UNESCO Recognition

On July 6, 2019, Piram Seowon and eight other academies were inscribed on the UNESCO World Heritage List as “Seowon, Korean Neo-Confucian Academies.” The serial property consists of Sosu, Namgye, Oksan, Dosan, Piram, Dodong, Byeongsan, Museong, and Donam Seowon. UNESCO recognized them as representative examples of Joseon Neo-Confucian academies established from the mid-sixteenth to the mid-seventeenth centuries and as exceptional testimony to the development of Neo-Confucian cultural traditions within Korean local communities.

Although the nine academies share common functions, each reflects a different scholar, region, and landscape. Piram Seowon represents the growth of Neo-Confucianism in Honam and the moral legacy of Kim In-hu. Architecturally, it is distinguished by its relatively level site, strong central axis, symmetrical organization, and the orientation of educational buildings toward the shrine.

The principal buildings extend along a north-south axis from Hwagyeonnu Pavilion to Cheongjeoldang, the dormitories, the inner gate, and Udongsa Shrine, forming the characteristic arrangement of education in front and ritual behind. UNESCO recognized not merely old wooden structures. It recognized a complete cultural system in which learning, ritual, community, architecture, landscape, and memory were integrated.

18. An Academy Close to Village Life

The landscape of Piram Seowon differs from that of academies located in dramatic mountain settings. It stands in a comparatively open and level environment close to fields and villages. This setting reflects the public character of the academy.

Learning was not intended only for people who withdrew from society. It was meant to educate a region, preserve shared memory, and influence community ethics. The complex is organized through symmetry and a strong central axis. This arrangement expresses the Neo-Confucian ideals of balance, order, and disciplined conduct.

Yet the purpose of order was not mechanical control. It was the careful alignment of mind, action, and responsibility. At Hwagyeonnu, the mind opens. At Cheongjeoldang, learning begins. At Jindeokjae and Sung-uijae, conduct is cultivated.

At Udongsa, the learner remembers the example of the teacher. The movement through the academy becomes a moral journey from selfishness toward impartiality, from knowledge toward practice, and from private desire toward communal responsibility.

19. Piram Seowon in the Age of Artificial Intelligence

We now live in a world of knowledge that Kim In-hu could never have imagined. Artificial intelligence can read and summarize vast numbers of documents, answer questions, generate images, and compose texts. The speed of knowledge production has expanded beyond any previous historical period. Yet Piram Seowon asks a question that comes before knowledge: With what kind of mind will knowledge be used?

For Kim In-hu, scholarship was not the possession of information. It was the power to make the mind impartial, restrain desire, preserve trust, and remain responsible to moral principle. Artificial intelligence can analyze Kim’s writings and organize the history of Piram Seowon. It cannot decide to sacrifice personal advantage for integrity.

Calculating what appears right is not the same as living according to what one believes is right. Machines can assist judgment. They cannot embody responsibility.

Education in the age of artificial intelligence therefore requires more than technological competence. It requires the ability to verify information, resist prejudice, respect others, accept responsibility for consequences, and maintain ethical standards when convenience encourages compromise. The impartiality represented by Hwagyeonnu and the pure integrity symbolized by Cheongjeoldang remain necessary human standards.

20. The Brush Is Soft, but the Principle Is Firm

The buildings of Piram Seowon do not proclaim themselves loudly. Passing through Hwagyeonnu, facing Cheongjeoldang, walking between Jindeokjae and Sung-uijae, and approaching Udongsa, one encounters the life of a scholar expressed through space.

First, the mind is opened. Then learning begins. Virtue is practiced in daily life. Finally, the learner stands before the memory of the teacher. This is the path of the seonbi preserved at Piram Seowon.

Kim In-hu did not leave his name through a long career of political power. He remained loyal to a king whose reign had ended almost as soon as it began. He withdrew from a corrupt political order and protected learning, teaching, and moral responsibility in his hometown. After his death, the scholars of Honam gave architectural form to his memory.

When the academy burned, they rebuilt it. When its site became unsafe, they moved it. They preserved books, woodblocks, documents, and rituals across generations. The brush of Piram still writes today.

It does not write only with ink upon paper. It writes within the human heart. The brush may be soft, but the principle it records remains firm like rock.

Political power changes. Institutions disappear. Names gradually fade. Yet the desire to live rightly does not vanish while someone continues to remember it.

Piram Seowon still asks: Will we become people who know much? Or will we become people who live rightly according to what we know?

The path of the seonbi is not merely a road into the past. In an age overflowing with information but easily deprived of moral direction, it remains a path that humanity must walk again.

Preview of Part 9

Episode 9 | Choe Chi-won and Museong Seowon — A Philosophy of Harmony

About the Author

Jung Yeon-Tae, who writes under the pen name Manri Pungchwi, is a cultural thinker and policy strategist. His research and writing connect history, philosophy, science and technology, interpreting Korea’s traditional culture as a source of future value in an era of civilizational change and artificial intelligence.

Read the Korean version

The Korean tobacco company said Wednesday that it held an opening ceremony for the new plant in Pasuruan, East Java, on October 6. KT&G President and CEO Bang Kyung-man attended the event along with Indonesian government officials and Yoon Soon-gu, South Korea’s ambassador to Indonesia.

The new facility is designed to accommodate up to nine cigarette production lines and eventually reach an annual capacity of 21 billion cigarettes. Combined with the existing Indonesian plant’s capacity of 14 billion cigarettes, KT&G’s total production capacity in Indonesia will rise to as much as 35 billion cigarettes a year.

The expansion is expected to make Indonesia KT&G’s largest overseas production base and a key hub for supplying both the domestic Indonesian market and international markets.

Products manufactured at the new facility will be shipped not only within Indonesia but also to markets including Taiwan, Mongolia, Nigeria and India. KT&G plans to capitalize on Indonesia’s strategic location between the Pacific and Indian oceans to improve supply efficiency across the Asia-Pacific region and other overseas markets.

KT&G entered Indonesia in 2011 through the acquisition of a local tobacco company. Since then, it has expanded its presence with brands such as Esse and Juara, a brand tailored to the Indonesian market. The company currently ranks fourth in Indonesia’s cigarette market and has expanded local manufacturing capacity in response to growing demand.

Construction of the new facility began in April 2024, followed by trial operations in the first half of this year. Its launch marks another major step in KT&G’s efforts to establish a geographically diversified overseas manufacturing system.

The company has invested approximately 2.4 trillion won in overseas production facilities as part of its global growth strategy. Following the completion of a new plant in Kazakhstan last year, the Indonesian facility brings KT&G closer to completing a large-scale investment cycle aimed at strengthening its international manufacturing infrastructure.

KT&G expects its production facilities in Indonesia, Russia, Kazakhstan and Türkiye to eventually have a combined maximum capacity of approximately 65 billion cigarettes annually. The company is targeting an overseas production ratio of more than 60% by 2028.

Industry observers say the expanded production footprint could give KT&G greater flexibility in responding to regional demand while reducing logistics costs and supply-chain dependence on individual manufacturing locations. The strategy also reflects the broader shift among global tobacco companies toward establishing production facilities closer to key consumer markets.

With the major overseas investment program entering its final phase, KT&G is also looking to translate improved production efficiency and cash generation into higher shareholder returns. The company has said it will continue measures such as share buybacks, cancellations and dividend expansion.

Bang Kyung-man said the launch of the Indonesian plant represents a major milestone in KT&G’s transformation into a global company.

“By bringing the Indonesian facility online, we have established an important foundation for our global production network,” Bang said. “We will maximize the efficiency of our overseas manufacturing bases, strengthen profitability and continue enhancing corporate and shareholder value.”

The company held a completion ceremony on October 6 at the Hombre Muerto salt lake in Argentina, located at an elevation of around 4,000 meters. POSCO Group Chairman Chang In-hwa and senior Argentine government officials attended the event, highlighting the strategic importance of the project to both sides.

The newly completed upstream facility represents a major step in POSCO’s long-term lithium strategy. With the second plant now adding 23,000 tons of annual lithium carbonate capacity, POSCO’s combined production capacity at Hombre Muerto has reached 48,000 tons per year, including the 25,000-ton capacity of its first plant.

POSCO Holdings invested approximately $954 million in the second facility. The lithium carbonate produced there is scheduled to be processed into battery-grade lithium hydroxide at a downstream facility being built in Korea’s Yulchon Industrial Complex in Jeonnam Province.

The Korean downstream plant, scheduled for completion in 2027, will have an annual production capacity of 25,000 tons of lithium hydroxide. When the two facilities are fully integrated, POSCO expects to establish a 50,000-ton annual production system based on lithium brine.

The project marks another milestone in POSCO Group’s efforts to secure critical battery minerals at the source rather than relying solely on external suppliers.

POSCO began developing its own lithium extraction technology in 2010, anticipating growing demand for lithium as electric vehicles and rechargeable batteries expanded. The group acquired mining rights to the Hombre Muerto salt lake in 2018 and completed its first commercial plant in 2024.

The latest investment also strengthens POSCO’s position in Argentina as the country seeks to attract large-scale investment in its mining and resources sector. POSCO Argentina secured approval under the government’s Regimen de Incentivo para Grandes Inversiones (RIGI) program in July, becoming the first Korean company to receive the investment incentives. The approval provides benefits including tax and customs-related support.

POSCO Argentina subsequently secured a $700 million short-term borrowing facility from IDB Invest in August, providing additional financial flexibility for its operations in the country.

POSCO is not stopping at the second plant. The company plans to accelerate preparations for its third and fourth lithium plants, each designed to produce 25,000 tons of lithium carbonate annually. The expansion would support POSCO’s goal of reaching 100,000 tons of annual lithium brine production by 2033.

The additional capacity is expected to position the group to capture growing demand from two areas of the battery market: lithium iron phosphate (LFP) batteries and energy storage systems (ESS), the latter of which is expected to benefit from the rapid expansion of AI data centers and their growing electricity requirements.

Rather than focusing exclusively on lithium hydroxide, POSCO plans to build a more balanced product portfolio covering both lithium hydroxide and lithium carbonate.

On the day of the completion ceremony, POSCO Argentina also signed memorandums of understanding with the governments of Salta and Catamarca provinces regarding the development of the third and fourth plants. The provincial governments agreed to support administrative procedures and permitting required for the projects.

Chairman Chang inspected the second plant during his visit and reviewed plans for the next phase of expansion with local officials. He also emphasized the importance of turning resource development into a stable source of production and earnings.

“By securing competitive resources ahead of demand and connecting them to production and profitability, we will strengthen the supply of critical minerals needed by Korean industry,” Chang said.

The expansion forms part of POSCO Group’s broader “Triple-Core” strategy, which identifies industrial resources, strategic resources and energy resources as key growth pillars.

For POSCO, the Argentine lithium business is therefore more than an overseas mining project. It represents an effort to build an integrated supply chain spanning resource acquisition, extraction, processing and battery-material production, while reducing Korea’s exposure to potential disruptions in the global critical-minerals market.

With the second plant now entering the final stage toward full-scale operations, POSCO Holdings is expected to focus on stabilizing production while laying the groundwork for its next round of expansion in Argentina.

PARAN LABEL was launched in South Korea in February 2025 with a focus on what Paris Baguette describes as a new standard for healthy bread. The brand combines proprietary fermentation techniques, health-conscious ingredients and balanced nutritional formulations while maintaining the taste and texture expected from bakery products.

The Singapore rollout follows an initial test at the brand’s Paragon store in February. After monitoring customer preferences and product performance, Paris Baguette expanded the range to all 28 stores across the country.

Seven PARAN LABEL products are currently available, including five bakery items and two sandwiches.

The expansion has already translated into stronger sales. Average 30-day sales in the bread category rose 57% compared with the average recorded from May through July, before the brand was rolled out across the entire Singapore network. Sales in the sandwich and salad categories also increased by 10%.

Among the best-selling products are the Smoked Salmon Ricotta Sandwich, Roasted Chicken Ricotta Sandwich and Multigrain Rye Bread. The rye bread has attracted repeat purchases for its nutty flavor and firm texture, while the smoked salmon sandwich has gained attention for combining smoked salmon, ricotta cheese and capers in a balanced meal-style offering.

The results point to a broader shift in consumer demand, with Singaporean customers increasingly looking for food that combines health considerations with taste and convenience. Paris Baguette has responded by incorporating attributes such as high fiber, reduced sugar, high protein and probiotics into selected products, while maintaining a focus on flavor and texture.

The company has also positioned PARAN LABEL as more than a traditional bread line. By offering sandwiches and salads alongside bakery products, it is seeking to make healthy bakery food part of consumers’ everyday meal choices rather than limiting it to snacks.

Local market accessibility has been another part of the strategy. Paris Baguette completed MUIS halal certification for all 28 Singapore stores this year, allowing Muslim consumers as well as other customers to access a wider range of products under halal-certified operations.

The company is also using the growing popularity of Korean culture in Southeast Asia to strengthen its bakery business. Rather than localizing the brand name, Paris Baguette has retained the Korean name “PARAN LABEL” in Singapore, reinforcing its Korean identity.

The company has recently expanded its “ANNYEONG” campaign across Southeast Asia to introduce Korean food and bakery culture to local consumers, incorporating Korean elements into both its products and marketing activities.

Industry observers say Singapore offers an attractive testing ground for Korean healthy bakery concepts because of its diverse consumer base and growing interest in health-conscious and convenient food options. The sharp increase in bread sales following the full-store rollout could provide Paris Baguette with a useful reference point as it considers further overseas expansion.

PARAN LABEL has also continued to grow in South Korea. The brand has surpassed 34 million units in cumulative sales since its launch and has expanded beyond healthy bread into cakes, sandwiches, desserts and beverages. Its latest offerings include five new healthy bread products featuring buckwheat, rice and rye.

A Paris Baguette representative said the company plans to use the Singapore rollout as a foundation for expanding PARAN LABEL into additional overseas markets, including the United States, with the goal of developing it into a global healthy bakery brand.

The Singapore expansion marks a significant step in Paris Baguette’s effort to turn a Korean health-focused bakery concept into an international business. With Korean food culture continuing to gain traction across Asia, the company is betting that the next K-Food category to travel overseas may be healthy bakery.

Nongshim brought the iconic red Shin Ramyun cup to the River Thames on Oct. 6 as part of a new brand campaign aimed at introducing the Korean instant noodle brand to a wider audience in Britain.

Standing about six meters tall, the installation set off from Royal Docks in the afternoon before making its way along the Thames toward the London Eye and Big Ben. Along the route, it passed some of the capital’s most heavily visited areas, including Tower Bridge and Southbank, where tens of millions of visitors and Londoners converge each year.

Under the campaign message “Work done. Day done. SHIN ME,” Nongshim is positioning Shin Ramyun not simply as a quick meal, but as a small moment of personal relaxation after a long day.

The concept was hard to miss on the river.

The oversized cup was illuminated for the evening and fitted with giant chopsticks. Steam rising from the installation was designed to make it look as though a freshly prepared bowl of hot Shin Ramyun was sitting inside the enormous cup.

The visual spectacle appeared designed for a city where public spaces routinely double as stages for global brands. Against the London skyline, the bright red Shin Ramyun cup offered a distinctly Korean image amid the historic architecture along the river.

For Nongshim, the campaign is part of a broader effort to deepen its presence in the UK, where Korean food and culture have gained greater visibility in recent years.

The rear of the installation also promoted Nongshim’s ongoing “SHIN TO WIN” consumer campaign in Britain. The promotion features scratch-off coupons inserted into Shin Ramyun multipacks sold through major retailers including Tesco, B&M and Sainsbury’s, with cash prizes available to winners.

The promotion runs from Aug. 10 through Oct. 31.

The Thames activation also reflects how Korean food brands are increasingly using experiences and visual storytelling rather than conventional advertising to connect with consumers overseas.

Rather than presenting Shin Ramyun solely as a Korean food product, Nongshim is attempting to attach the brand to a familiar everyday ritual: switching off after work, finding a quiet moment and enjoying something comforting.

That message is particularly suited to London, where the Thames serves both as a backdrop for major tourist attractions and as part of the daily landscape for millions of residents.

“Nongshim hopes to use this campaign to offer British consumers a new brand experience centered on enjoying their own moment of relaxation with Shin Ramyun after a busy day,” a Nongshim official said.

As the giant red cup moved along the Thames, it became more than an advertisement for instant noodles. For a few hours, it turned one of London’s most familiar waterways into a moving showcase for a distinctly Korean brand — and a simple invitation to call the working day over and make time for a bowl of something spicy.

The automakers showcased their future mobility vision and key technologies at the Aachen Automotive Colloquium 2026, held in Aachen, Germany, from Oct. 5 to 7.

A key highlight was the companies’ first keynote address at the event since the colloquium was launched in 1987. Manfred Harrer, head of R&D at Hyundai Motor and Kia, delivered the keynote under the theme “Beyond Boundaries, Expanding a Sustainable Future.”

Harrer outlined Hyundai Motor and Kia’s multi-energy strategy, emphasizing that different electrification technologies can serve different markets and customer needs.

The approach reflects a changing global automotive market. While the long-term shift toward electrification remains intact, the pace of EV adoption and charging infrastructure development varies significantly by region. Automakers are therefore increasingly maintaining a broader range of powertrain technologies to respond to different market conditions.

Hyundai Motor and Kia also placed artificial intelligence (AI) at the center of their future R&D strategy.

Harrer said AI should not be viewed as a replacement for engineers, but as a tool that can expand their capabilities and enhance creativity. The companies are working to incorporate AI and digital technologies into vehicle development to improve engineering efficiency and accelerate innovation.

Digital twins and virtual validation are also being used to transform the development process. By simulating vehicle performance and various operating conditions before physical prototypes are built, Hyundai Motor and Kia aim to shorten development cycles and improve product quality.

The companies also presented a vision in which vehicles continue to evolve after they are delivered to customers.

Vehicles could learn individual driving patterns and preferences and use that data to provide more personalized driving assistance. Software updates and accumulated vehicle data would then feed into further improvements in products and services.

Hyundai Motor and Kia refer to this approach as a “Data Flywheel,” in which vehicle and customer data are continuously fed back into product and service development.

Industry observers say the strategy reflects the broader transformation taking place in the automotive sector. Competition is increasingly moving beyond traditional hardware such as engines and transmissions toward the integration of electrification, batteries, software, AI and data.

The automakers also used the Aachen event to demonstrate their technological capabilities through an extensive exhibition.

Hydrogen-related technologies on display included a proton exchange membrane (PEM) electrolyzer stack, hydrogen storage tanks and the fuel-cell system used in the all-new NEXO. Electrification technologies included battery packs and cells, a two-stage motor system, the next-generation hybrid transmission system TMED-II and power modules.

Visitors were also able to experience Hyundai Motor and Kia’s latest electrified vehicles through test drives of the all-new NEXO and the GV60 Magma. Participants drove designated routes of approximately 11 to 12 kilometers near the event venue.

The automakers are also scheduled to participate in technical sessions covering hydrogen, electrification and batteries on the final day of the conference. Topics include the future of fuel-cell technology, strategies for expanding xEV technologies and next-generation battery-based mobility.

Hyundai Motor and Kia’s expanded presence at the Aachen Colloquium is significant as the companies seek to strengthen their technology credentials in Europe. Their decision to showcase multiple powertrain technologies alongside AI and digital development capabilities suggests a strategy focused not only on electrification itself, but also on flexibility and technological integration.

The Aachen Automotive Colloquium, organized by a research institute affiliated with RWTH Aachen University, is one of Europe’s leading automotive technology conferences. Automakers, suppliers, universities and research institutions participate in the annual event to discuss emerging vehicle technologies and the future direction of the automotive industry.

The Korean power equipment and energy solutions provider said on Oct. 7 that it had signed an agreement with Taiwan-based Nan Ya Plastics and KHI to establish a joint venture dedicated to power business in Taiwan.

The new company, named Nan Ya Le Xing (南亞樂星), will be jointly owned by LS Electric, Nan Ya Plastics and KHI. LS Electric will hold a 50% stake, while Nan Ya Plastics and KHI will own 40% and 10%, respectively.

The partnership is designed to transform LS Electric’s position in Taiwan from a supplier of power equipment into a more comprehensive project partner capable of handling engineering, bidding and technical support locally.

The joint venture will initially focus on securing power infrastructure projects in Taiwan and building the capabilities needed to respond quickly to local customers. LS Electric’s expertise in power equipment manufacturing and engineering will be combined with Nan Ya Plastics’ local manufacturing infrastructure and EPC network.

The move comes as Taiwan accelerates investment in its electricity infrastructure amid the country’s transition toward renewable energy.

Taiwan is pursuing a carbon-neutrality strategy that calls for renewable energy to account for as much as 70% of its overall power mix by 2050. The transition is expected to drive demand for power infrastructure capable of integrating renewable generation while maintaining grid stability.

Investment opportunities are emerging across areas including environmentally friendly power equipment, energy storage systems and digitally enabled smart grids. For power equipment companies, the market is increasingly shifting from individual equipment supply toward integrated solutions and infrastructure projects.

LS Electric’s new joint venture is positioned to capitalize on that shift by combining its technology with local market access.

Nan Ya Plastics, a key affiliate of Formosa Plastics Group, is Taiwan’s largest petrochemical and electronic materials and products company. The company has also been expanding its power solutions business, moving beyond distribution equipment into areas such as extra-high-voltage transmission equipment and power grid stabilization.

KHI, another partner in the venture, was established through an investment by Formosa Plastics Group to commercialize promising technologies developed through research with KAIST. KHI has invested in the joint venture based on its assessment of LS Electric’s competitiveness and growth potential in the global power industry.

The new venture builds on a memorandum of understanding signed by LS Electric and Nan Ya Plastics in March last year covering cooperation in environmentally friendly power equipment and energy-efficiency solutions.

Their partnership had initially centered on the supply of power equipment. The establishment of the joint venture expands the relationship into engineering and project-based business, allowing the companies to pursue local contracts together.

Industry observers say the partnership reflects LS Electric’s intention to make Taiwan a strategic market for its power infrastructure business rather than simply an overseas sales destination.

The joint venture could also give LS Electric greater access to local projects and improve its ability to respond to customer requirements by combining Korean technology with Taiwan-based production and EPC capabilities.

An LS Electric official said the company expects the joint venture to elevate its Taiwan business from equipment supply to engineering-based project contracts.

“We expect the combination of the two companies’ capabilities to enable us to achieve early results in Taiwan’s power infrastructure project market,” the official said.

Daewoong Pharmaceutical said on Oct. 7 that it successfully completed the Industrial Skills Event (ISE), an industry-academia collaboration program conducted with the Korean National Association for Pharmaceutical Students (KNAPS). A total of 80 students participated, including 48 pharmacy students from Korea and 32 students from the Indonesian pharmacy student organization BEM Fakultas Farmasi Universitas Indonesia (BEM FF UI).

The central assignment was to develop a market-entry strategy for Daewoong Pharmaceutical’s pain-relief brand EasyN6 Eve in Indonesia.

The project began with online workshops in which Korean and Indonesian students worked together in teams. Participants examined the Indonesian pharmaceutical market and consumer needs before developing strategies covering product positioning, distribution and marketing.

The program culminated in team presentations, with Daewoong Pharmaceutical employees serving as mentors and judges. The company’s working-level staff provided feedback on the feasibility and practical application of each proposal from an industry perspective.

The winning team focused on women experiencing menstrual pain and related discomfort, including bloating and abdominal swelling. The students proposed a targeted positioning strategy combined with a distribution model linking local pharmacies with digital channels.

The project highlights a growing emphasis within the pharmaceutical industry on practical global-market experience for future talent. Rather than limiting the program to classroom-based learning, students were asked to tackle a real-world market-entry challenge and develop a strategy from consumer research through commercialization.

The participation of Indonesian pharmacy students also added a local perspective to the project. Their understanding of local consumers and market conditions helped Korean participants identify factors that could be overlooked when approaching the Indonesian market from Korea alone.

Industry observers say such cross-border industry-academia programs can provide value beyond talent development. For pharmaceutical companies expanding overseas, access to local market insights and professionals capable of translating those insights into commercial strategies is becoming increasingly important.

The program also provided Daewoong Pharmaceutical with an opportunity to engage directly with young talent interested in the global pharmaceutical industry. As Korean pharmaceutical companies accelerate their international expansion, collaboration with universities and student organizations is expected to become an increasingly important tool for developing globally oriented talent.

The winning team, consisting of pharmacy students Kim Yoon-seo, Kim Min-joo and Kim Min-jung from Ewha Womans University, said the project was particularly meaningful because they were able to identify the core value of EasyN6 Eve based on Indonesian consumer needs and consider how the product could establish itself in the local market.

Park Eun-kyung, head of Daewoong Pharmaceutical’s CH Division, said the company was impressed by the creative ideas generated through collaboration between students from the two countries.

“We were able to see the strong potential of young talent through the fresh ideas developed from the perspectives of students in both countries,” Park said. “We will continue to expand industry-academia collaboration opportunities so that young people can build practical capabilities needed in the global pharmaceutical industry.”

The South Korean electronic components maker said Tuesday that it has signed a contract worth approximately 290 billion won ($200 million) with a major global customer. The agreement covers supplies from January through December 2027.

The deal marks the company’s sixth long-term supply agreement disclosed this year and adds to a series of contracts targeting components used in next-generation computing infrastructure.

MLCCs play a critical role in electronic systems by stabilizing power supplied to semiconductors and reducing signal interference between components. AI servers place particularly demanding requirements on these components because they operate continuously under high-temperature and high-voltage conditions.

As AI workloads become increasingly compute-intensive, the amount of MLCCs installed in servers is also rising sharply. An AI server rack can require up to 600,000 MLCCs, more than 10 times the amount used in conventional servers, according to Samsung Electro-Mechanics.

The company has been focusing on high-capacity, high-reliability MLCCs designed for AI servers while expanding production capabilities to accommodate growing demand from global customers.

Samsung Electro-Mechanics has signed long-term agreements covering MLCCs, silicon capacitors and inductors with multiple multinational customers. Since May, the company has announced contracts totaling about 3.9 trillion won, according to its regulatory disclosures.

The agreements are expected to provide a more stable demand base while helping Samsung Electro-Mechanics accelerate its shift toward higher-value electronic components.

The AI server MLCC market is also expected to expand rapidly. Goldman Sachs estimates that the market will grow from $1.4 billion in 2025 to $5.8 billion by 2030, representing an annual growth rate of roughly 34%.

Samsung Electro-Mechanics currently holds more than 40% of the global AI server MLCC market, putting it among the leading suppliers as hyperscalers and data center operators continue to increase AI infrastructure spending.

The company said it plans to pursue additional supply opportunities with global customers while expanding the share of AI-focused products in its portfolio.

“As the AI market continues to grow, customers are demanding increasingly sophisticated MLCCs,” a Samsung Electro-Mechanics official said. “We will strengthen our competitiveness in the global market by expanding AI-related products based on differentiated technology and reliable supply capabilities.”

The game went live on October 5 local time across global markets through Steam and NC’s PURPLE platform. The strong opening follows an early access period that began on September 30 for players who purchased the Founder’s Pack.

Aion 2 gained momentum ahead of its full launch, remaining at the top of Steam’s global Top Sellers chart throughout the five-day early access period. The game also reached the No. 1 position in the chart in 15 individual countries.

Its performance has continued following the official launch. As of October 6, Aion 2 ranked first on Steam’s “Popular New Releases” and “Trending Free” charts, underscoring strong early interest from players across international markets.

NC is positioning Aion 2 as a major title in its global expansion strategy. The game is available across four regions — North America, South America, Europe and Japan — and supports eight languages, including English, Japanese, French, German and Spanish.

To coincide with the global launch, NC has also released a behind-the-scenes documentary featuring the development team. The video offers a look at the studio’s development goals and the process behind creating Aion 2.

The latest launch marks an important test of NC’s ability to build a global audience for its MMORPG portfolio. With the title already showing strong performance across key Steam indicators, attention is now turning to whether Aion 2 can sustain its momentum and establish a long-term presence in overseas markets.

Further information on Aion 2, including game content and updates, is available through the title’s official website.

The financial and accountability structure surrounding artificial intelligence is becoming more complex. Wall Street is arranging $60 billion to support Anthropic’s access to AI chips, a U.S. startup has introduced a large open-weight model that companies can operate themselves, and the New York City Council has called major AI companies and former researchers to testify about safety measures and responsibility for loss of control. Taken together, the developments show that investment, operational and safety risks once concentrated largely around AI developers are increasingly being shared by financial institutions, enterprise users and public authorities.

$60 Billion in Financing for AI Chips

Wall Street firms including Bank of America, Citigroup and Morgan Stanley have begun marketing a $60 billion financing package to support Anthropic’s leasing of Google AI chips. The structure reportedly includes $42 billion in senior secured debt, backed in part by residual-value support from Broadcom, and another $18 billion in junior debt. Blackstone has committed $9 billion of the junior portion.

The significance goes beyond the headline number. Securing AI compute is becoming too capital-intensive to rely solely on corporate cash and conventional bond issuance, bringing bank loans, guarantees, leasing structures and private capital deeper into AI infrastructure financing. Anthropic co-founder and CEO Dario Amodei has warned that if companies misjudge future AI demand, enormous capital spending could result in a “painful adjustment.”

That changes the economics of the AI boom. If projected demand fails to materialize, losses may no longer remain concentrated within model developers. Lenders, investors and infrastructure providers could also be exposed to the consequences of an industry built on exceptionally large assumptions about future computing demand.

The U.S. Re-enters the Open-Model Race

NVIDIA-backed U.S. AI startup Reflection AI unveiled Beam, its first open-weight model, on October 5. Beam uses a mixture-of-experts architecture with 501 billion total parameters and 23 billion activated for each task, with a focus on coding, reasoning and agentic workloads.

Reflection AI is positioning Beam as a competitor to leading Chinese open-weight models and is conducting final safety evaluations and validation. Co-founder and CEO Misha Laskin has framed the effort around restoring U.S. leadership in open intelligence.

Beam’s competitive standing will require further independent evaluation. More consequential, however, may be how such models are deployed. With closed models, developers largely control access and operation. Open-weight models give companies and developers considerably more ability to deploy and operate models within their own systems.

As Chinese companies have built influence in the open-weight ecosystem, the U.S. response is becoming strategically significant. The contest is therefore not only about which model performs best on benchmarks, but which models become embedded in the systems that companies and developers actually use.

New York Scrutinizes AI Safety Measures

The New York City Council held a full committee hearing on AI safety on October 5, bringing representatives from OpenAI, Anthropic, Google and Meta together with former AI researchers to address advanced AI risks, corporate safeguards and responsibility for potential failures of control. Safety standards that AI companies have largely designed and assessed internally are increasingly facing public scrutiny.

Jacob Coxon, a former researcher at OpenAI and Anthropic, warned at the hearing about the possibility that continued advances could eventually produce AI systems that humans struggle to control. His testimony raised a broader concern over whether safeguards designed and evaluated primarily inside AI companies are sufficient.

Such scenarios do not represent a consensus across the AI research community. The more immediate development is institutional: policymakers are examining corporate safety practices and considering measures including independent evaluation and whistleblower protections. AI safety is becoming an issue of external verification and accountability rather than solely an internal engineering decision.

The three developments reveal a deeper change in who carries the costs and risks of AI. Banks and private capital are financing Anthropic’s compute infrastructure, companies and developers can assume greater responsibility for operating open-weight models, and public authorities are beginning to scrutinize safety standards previously managed largely within AI companies.

As that structure develops, the benefits and liabilities of AI are distributed across more participants. If demand falls short of expectations, lenders and investors face financial exposure. As open-weight deployment grows, enterprises take on greater security and operational responsibility. And when serious failures occur, determining responsibility solely according to a developer’s internal standards becomes increasingly difficult.

The numbers that matter, therefore, are no longer limited to parameter counts or benchmark scores. Who captures the returns on AI investment, who absorbs the losses when demand forecasts fail, and who is accountable when systems cause harm are becoming measures of the industry’s underlying economics.

The South Korean chemical company said Tuesday that it will showcase a range of next-generation engineering materials at Fakuma 2026, a major plastics industry trade fair taking place in Friedrichshafen, Germany, from Oct. 12 to 16.

LG Chem plans to use the event to highlight material solutions designed for automotive, electrical and electronic applications, as well as AI data centers, robotics and high-performance breathable films.

Rather than focusing solely on displaying individual products, the company will emphasize customized solutions aimed at addressing specific challenges faced by industrial customers. Its approximately 70-square-meter exhibition space will be organized around three themes: cost efficiency, sustainability and innovation.

Under the cost-efficiency theme, LG Chem will present paint-free materials for automotive applications and solutions designed to replace metal components. Such materials can help manufacturers reduce vehicle weight and improve production efficiency.

The sustainability section will feature bio-based materials and low-carbon solutions, reflecting growing demand in Europe for products that can help manufacturers meet increasingly stringent environmental requirements.

LG Chem will also highlight advanced engineering plastics developed for AI data centers and robotics. Engineering plastics are increasingly being used as alternatives to metal because they can contribute to lighter components and improved energy efficiency while offering high levels of durability and performance.

Europe is a key target market for these materials as the region accelerates vehicle electrification and the transition toward more environmentally sustainable manufacturing. LG Chem aims to capitalize on rising demand for lightweight, high-performance materials by presenting solutions tailored to emerging industrial applications.

The company will also use the exhibition to deepen technical cooperation with customers. More than half of its booth will be dedicated to meeting areas where LG Chem plans to hold technical consultations and discuss potential joint development projects.

The company will further introduce its proprietary SISO solution, which allows customers to predict and verify material performance during the product design stage. LG Chem expects the platform to help customers evaluate materials earlier in the development process and accelerate collaboration.

“Customers are looking for more than materials themselves; they need next-generation solutions that can turn future possibilities into reality,” said Lee Chang-hyun, head of LG Chem’s Engineering Materials Business. “We will stay ahead of customer needs and propose differentiated material solutions that create new growth opportunities and value.”

Through its participation in Fakuma 2026, LG Chem aims to strengthen its presence in Europe while expanding partnerships across automotive, electronics, AI infrastructure and robotics markets.

The company said on Oct. 6 that two of its proposals have been selected for government-backed research projects led by the Ministry of Trade, Industry and Energy and the Korea Planning & Evaluation Institute of Industrial Technology.

The projects cover the development of next-generation collaborative robots based on on-device AI semiconductors and an intelligent welding solution combining collaborative robots with artificial intelligence.

Their combined research and development budget is approximately 98.9 billion won ($70 million), including about 68.1 billion won in government funding.

Doosan Robotics plans to use the first project to develop core technologies that enhance robots’ perception, decision-making and control capabilities, while the second will apply AI-powered robotics to complex manufacturing processes.

The company ultimately aims to integrate the technologies into what it calls a “Korean physical AI collaborative robot platform.”

Collaborative Robots Designed to Think Without the Cloud

The first project focuses on developing next-generation collaborative robots capable of perceiving their surroundings, making decisions and controlling their movements without relying on external PCs or cloud computing.

The project will integrate an AI neural processing unit, or NPU, and real-time communication and control functions into a Korean-developed AI system-on-chip, or SoC, for use in collaborative robots.

Using data from cameras, tactile sensors, force and torque sensors, and voice inputs, the robots will be designed to process information and make decisions directly at industrial sites.

Doosan Robotics plans to develop three collaborative robot models with payload capacities of 5 kilograms, 10 kilograms and 20 kilograms. It will also develop an industrial humanoid capable of dual-arm operation and omnidirectional movement.

The robots will undergo 72-hour continuous-operation demonstrations at two or more customer sites as well as international safety certification. Doosan Robotics aims to commercialize them as standard products beginning in 2031.

On-device processing could also broaden the use of collaborative robots in security-sensitive industries such as nuclear power, defense and biotechnology, where concerns over communication latency or transferring industrial data to external cloud systems can limit automation.

The 54-month project will be led by Doosan Robotics, with participation from Aidin Robotics, Mobilint, the Korea Electronics Technology Institute, the Korea Institute of Industrial Technology, SAIGE and industry-academic cooperation groups from several Korean universities.

AI to Learn Skilled Workers’ Welding Techniques

The second project focuses on developing an intelligent welding solution combining collaborative robots, AI and digital twin technology.

The system is designed to use cameras to recognize welding targets and automatically adjust the robot’s working position. AI will also learn from the movements of skilled welders to generate welding paths and weaving motions, in which a welding torch moves from side to side to achieve consistent weld width and quality.

Doosan Robotics aims to achieve more than 95% alignment between robot-generated welding paths and those of skilled workers, while limiting path deviation to within ±3 millimeters under testing by an accredited national institution.

The company also plans to automate multilayer welding of five or more layers and reduce working time by more than 50%.

Testing will take place at Doosan Enerbility’s Nuclear Power Center. The technology will initially be applied to processes such as tack welding, preheating and welding in the production of nuclear power equipment.

After securing the necessary nuclear quality certifications, Doosan Robotics plans to expand the technology into shipbuilding, industrial plants and defense manufacturing.

The project will run for 45 months, with Doosan Robotics serving as the lead organization. DEEPX, SAIGE and Changwon National University’s industry-academic cooperation foundation will participate as research partners, while Doosan Enerbility will participate as an end-user company.

Integrating AI Chips, Software and Robotics

The two projects are intended to integrate AI semiconductors, software and robotic hardware into a physical AI platform built around domestic technologies.

Rather than limiting robots to predefined repetitive tasks, Doosan Robotics is seeking to develop systems capable of perceiving changes in their surroundings, making decisions and reproducing complex skills traditionally dependent on experienced workers.

“Labor shortages and the aging of skilled workers in manufacturing are structural challenges that cannot be addressed simply by increasing the number of automated systems,” said Park In-won, CEO of Doosan Robotics.

“By combining on-device AI semiconductors with our collaborative robot platform, we aim to develop physical AI collaborative robots capable of learning and reproducing even the tacit knowledge of skilled workers and turn them into a new source of competitiveness for Korean manufacturing.”

The Korean pharmaceutical company said on Oct. 6 that it presented preclinical research on a small-molecule drug conjugate, or SMDC, targeting fibroblast activation protein (FAP) at Discovery on Target 2026, held in Boston from Sept. 28 to Oct. 1.

The approach is designed to address some of the limitations associated with antibody-drug conjugates, or ADCs, by using smaller molecules that can more effectively penetrate tumor tissue and reach targets within the tumor microenvironment.

Targeting FAP to Complement ADC Approaches

ADCs have emerged as a major class of anticancer therapeutics by delivering cytotoxic drugs to tumor-associated antigens expressed on the surface of cancer cells.

Their relatively large molecular size, typically around 150 kDa or more, however, can limit penetration and drug distribution within solid tumors. Access to cancer cells can be particularly challenging in tumor microenvironments densely populated with cancer-associated fibroblasts, or CAFs, and extracellular matrix components.

Hanmi Pharmaceutical has focused on SMDCs as a potential approach to complement these limitations. Because SMDCs are considerably smaller than antibody-based conjugates, the company is exploring their ability to penetrate tumor tissue and the surrounding microenvironment.

For its new modality, Hanmi selected FAP as the target.

FAP is highly expressed in CAFs within the tumor microenvironment but has limited expression in most normal adult tissues. High levels of FAP have also been observed across multiple solid tumors, including lung cancer, soft tissue sarcoma and gastrointestinal cancers such as pancreatic and colorectal cancer.

New Compound Identified Through DEL Screening

Hanmi Pharmaceutical used its DNA-Encoded Library, or DEL, screening technology to identify a novel class of small molecules that selectively bind to FAP.

The company then developed an SMDC by conjugating a cytotoxic payload to the FAP-targeting molecule.

The modality incorporates a linker designed to remain stable in the bloodstream and become selectively activated within the tumor microenvironment, allowing the drug payload to act at the tumor site.

In preclinical studies presented at the conference, the SMDC demonstrated antitumor activity in mouse models implanted with FAP-expressing cells.

Hanmi also reported strong tumor growth inhibition and prolonged survival in an orthotopic pancreatic cancer mouse model.

The company said the approach differs from conventional ADC strategies that rely on the expression of specific antigens on individual cancer cells. Instead, the SMDC targets FAP expressed by CAFs, a common component of the tumor microenvironment across a range of cancer types.

This could potentially broaden the therapeutic approach from directly targeting cancer cells to addressing the surrounding environment that supports tumor growth.

Research to Be Presented at MD Anderson Cancer Center

Hanmi Pharmaceutical is also presenting the findings at another international scientific meeting.

The research is scheduled for an oral presentation at the “Nature Conference – Next Generation Cancer Therapeutics,” being held from Oct. 6 to 8 at the MD Anderson Cancer Center in Texas.

The program reflects Hanmi Pharmaceutical’s broader effort to expand its research portfolio beyond metabolic diseases, including obesity, into next-generation oncology modalities.

“The SMDC modality is differentiated from conventional ADCs in that it can deliver drugs by targeting FAP expressed in cancer-associated fibroblasts within the tumor microenvironment, rather than depending on the expression of a specific antigen on individual cancer cells,” a Hanmi Pharmaceutical official said.

“The findings demonstrate the potential of this approach to expand the therapeutic target from cancer cells themselves to the surrounding tumor microenvironment,” the official added.

The certification follows regulatory approval through Japan’s Pharmaceuticals and Medical Devices Agency (PMDA) last month and provides an additional institutional pathway linking the software to Japan’s diagnostic imaging management and reimbursement framework.

With the JRS certification, medical institutions that meet the facility requirements for Diagnostic Imaging Management Add-ons 3 and 4 can use AVIEW Lung Nodule CAD within a framework that incorporates safety and quality management requirements for AI-assisted diagnostic imaging software.

The certification is expected to support the software’s integration into clinical imaging workflows and its continued use by Japanese medical institutions.

JRS Certification Follows PMDA Approval

Japan’s medical AI market requires a distinction between medical device authorization, professional society-based AI management standards and reimbursement.

PMDA-reviewed approval serves as the regulatory gateway for commercialization as a medical device in Japan. JRS certification, meanwhile, provides an institutional connection to the safety and quality management framework governing AI software used in actual diagnostic imaging practice.

JRS certification is separate from third-party medical device certification conducted by registered certification bodies.

Japan revised its medical reimbursement system in 2022 to incorporate appropriate safety management requirements for AI-based diagnostic imaging support software into facility standards for Diagnostic Imaging Management Add-on 3. The JRS subsequently established a certification system for AI software requiring safety and quality management in clinical imaging environments.

As of Oct. 1, 2026, 45 AI software products were included on the JRS official certification list.

Coreline Soft had previously obtained PMDA approval for AVIEW Lung Nodule CAD. According to the company’s research, it is the first Korean-developed CT-based lung nodule CAD product to receive such approval in Japan.

By securing both regulatory approval and JRS certification within roughly a month, Coreline Soft has moved beyond obtaining permission to commercialize the product and expanded the institutional foundation for its use in Japanese clinical settings.

M3 AI Distribution Expands Portfolio from Lung to Heart

Coreline Soft is simultaneously expanding another part of its Japanese business.

The company recently signed a supply agreement with M3 AI for AVIEW CAC, its cardiovascular imaging detection and diagnostic support software.

Its Japan strategy is therefore developing across three areas: regulatory entry through PMDA approval, connection to the reimbursement framework through JRS certification, and commercialization and distribution through M3 AI.

Japan provides a significant market for medical imaging AI because of its extensive CT infrastructure. The country has more than 115 CT scanners per million people, among the highest levels globally, while the number of physicians per 1,000 people stands at about 2.6, below the OECD average of 3.9.

The combination of high imaging capacity, limited medical personnel and a rapidly aging population is increasing the potential role of AI in supporting diagnostic imaging workflows.

Coreline Soft has also expanded its existing customer base in Japan. Following its acquisition of VUNO’s Japanese chest CT AI assets last year, the company has grown its network to approximately 80 customer hospitals within about a year.

The JRS certification adds another institutional foundation for expanding use of its software among existing and new customers.

Japan Moves Toward Wider LDCT Lung Cancer Screening

Changes in Japan’s lung cancer screening environment could provide another opportunity for medical imaging AI.

Japan’s National Cancer Center has given a Grade A recommendation to low-dose CT, or LDCT, lung cancer screening for high-risk smokers, while the Ministry of Health, Labour and Welfare is conducting demonstration projects in five municipalities this year.

If Japan expands LDCT screening and the use of AI in medical imaging, Coreline Soft plans to leverage its existing regulatory, institutional and distribution infrastructure to increase adoption and usage at hospitals.

The company is also expanding from lung imaging into cardiovascular analysis by offering both AVIEW Lung Nodule CAD and AVIEW CAC in Japan. The strategy is aimed at building a broader chest CT portfolio covering both pulmonary and cardiovascular conditions.

“Following PMDA regulatory approval, we have now secured certification that meets the requirements related to hospital reimbursement, providing clearer incentives for Japanese medical institutions to adopt and continuously use AI,” said Kim Jin-kook, CEO of Coreline Soft.

“In Japan, one of the world’s most CT-intensive healthcare markets, we will use this certification as a foundation to expand real-world utilization of our chest CT portfolio, covering areas from lung to cardiovascular imaging,” Kim said.

Celltrion said Tuesday that it launched Omlyclo (omalizumab) in the United States on Oct. 5 local time. The drug is used to treat conditions including chronic spontaneous urticaria and allergic asthma.

Omlyclo is the first omalizumab biosimilar launched in the U.S., giving Celltrion a first-mover advantage as it seeks to secure market share ahead of potential competitors.

The company has also secured an important foothold in the U.S. reimbursement system. Omlyclo was included as a preferred drug on the commercial formularies managed by Optum, one of the country’s major pharmacy benefit managers (PBMs), from the time of its launch.

PBM formulary placement is considered a critical factor in expanding access to prescription medicines in the U.S., as it can influence reimbursement conditions and patients’ out-of-pocket costs. Celltrion is also negotiating with other major PBMs, with some discussions reportedly nearing completion.

Omlyclo is priced at a wholesale acquisition cost approximately 15% below that of the reference product, giving Celltrion an additional lever to compete in the U.S. market.

The product is available in three dosage strengths — 75 mg, 150 mg and 300 mg — allowing physicians to select treatment options according to individual patient needs.

Omlyclo also offers broad regulatory coverage. The U.S. Food and Drug Administration approved the product in March 2025 for all indications held by the reference product in the U.S., including asthma, chronic rhinosinusitis with nasal polyps, chronic spontaneous urticaria and IgE-mediated food allergy.

Celltrion has also secured interchangeability status for Omlyclo. Subject to applicable state laws and insurance requirements, the designation allows pharmacists to substitute the biosimilar for the reference product in certain circumstances, potentially improving patient access and supporting wider adoption.

The U.S. launch follows strong performance for Omlyclo in Europe, where Celltrion has leveraged its first-mover position to rapidly expand market share.

Omlyclo is currently available in 26 European countries. In Spain, its market share surpassed 90% as of July, according to IQVIA. In Denmark, Celltrion expects its share to reach approximately 98% by the end of the year following successful participation in national tenders.

Celltrion plans to apply the same market-entry strategy in the U.S. — launching early, securing reimbursement access and then expanding prescriptions through its commercial infrastructure.

The company believes its established direct-sales network in the U.S. will give Omlyclo an additional advantage. Celltrion can leverage its existing sales, distribution and healthcare-provider networks built around its other biosimilar products.

Its existing portfolio has already gained significant traction in the U.S. Truxima, a rituximab biosimilar, held a 39% market share in August and remained the market leader for the sixth consecutive month. Inflectra, based on infliximab, recorded a 31% share and maintained a leading position.

Other Celltrion products are also gaining ground. Steqeyma, a ustekinumab biosimilar, and Yuflyma, an adalimumab biosimilar, recorded market shares of 15% and 12%, respectively.

Industry observers say the key challenge for Omlyclo will be translating its first-mover advantage and early PBM access into actual prescribing volume. In the U.S., formulary coverage, physician adoption and distribution access all play important roles in determining the speed at which a biosimilar can gain market share.

Celltrion plans to expand its PBM coverage while stepping up marketing efforts targeting healthcare providers, insurers and distribution channels. The company also intends to use real-world prescription data and feedback from physicians and patients to refine its U.S. commercial strategy.

“Omlyclo combines the first-mover advantage as the first omalizumab biosimilar in the U.S. with broad indications, interchangeability and three dosage strengths,” said Kim Bon-joong, CEO of Celltrion’s U.S. subsidiary.

“By securing preferred-drug status with a major PBM at launch, we have also established a solid foundation for expanding prescriptions from the outset,” he said. “We will combine the experience gained from our successful first-mover strategy in Europe with our direct-sales capabilities in the U.S. to develop Omlyclo into one of Celltrion’s key growth products in the American market.”

Moldova rose eight places to 82nd in StartupBlink’s 2026 Global Startup Ecosystem Index, according to an announcement issued by Startup Moldova Foundation in Chișinău on October 5. The country’s population-adjusted ecosystem score increased by 91.9% between the 2025 and 2026 editions, the highest increase in Europe and third-highest worldwide, the foundation said. The result also brought recognition at the Startup Ecosystem Awards 2026, where the foundation accepted an award on behalf of Moldova’s startup ecosystem.

Moldova ranked 21st in Eastern Europe, while its capital, Chișinău, placed 464th globally and 35th in the region. StartupBlink’s index assesses startup quantity, ecosystem quality and the business environment, covering more than 1,500 cities and 120 countries in its 2026 edition. The reported 91.9% increase refers to the index score, rather than a corresponding rise in startup numbers, revenue or investment.

According to the foundation, more than 300 startups operate in Moldova, and local technology startups attracted over $17 million in financing in 2025. That total included equity investment, convertible financing and grants, with funding from angel investors and international venture funds. The figures point to a financing base that extends beyond venture-capital equity alone.

Olga Melniciuc, executive director of Startup Moldova Foundation, said the country’s small domestic market had encouraged entrepreneurs to develop products for international customers. She pointed to technical talent, tax arrangements, financing instruments and investor networks as factors supporting startup development. The next challenge, she said, was to translate those conditions into internationally competitive products, larger investment rounds and exits, while retaining teams and intellectual property in Moldova.

The foundation cited two acquisitions in 2025 as examples of companies with Moldovan founders or development operations reaching international markets. Social media collaboration platform Planable was acquired by SE Ranking after growing to more than 7,000 customers worldwide. Founded by Moldovan entrepreneurs, Planable developed significant parts of its product and engineering operations in Chișinău. Parkopedia, founded by Moldovan-born entrepreneur Eugene Tsyrklevich and supported from its early stages by a development team in Chișinău, was acquired by Sweden’s EasyPark Group.

Technology tax framework supports an export-oriented sector

The Moldova Innovation Technology Park (MITP) provides a nationwide framework through which eligible resident companies can operate from locations across the country. According to the foundation’s announcement, residents pay a single tax of 7% of sales revenue that replaces several separate taxes and contributions, alongside simplified reporting requirements. The announcement states that the government has guaranteed the stability of the regime until December 31, 2035.

The framework also includes an IT Visa programme intended to simplify employment and residency procedures for foreign executives, investors and highly qualified technology specialists. These arrangements support companies seeking to recruit international staff. StartupBlink identified regulatory modernisation, EU-oriented reforms and technical talent among the characteristics of Moldova’s startup environment, according to the foundation.

MITP exceeded 3,000 active resident companies in September 2026, compared with 338 when it began operating in 2018. Its residents employ more than 27,000 people, with founders and investors from over 50 countries represented in the park. Approximately 90% of residents’ products and services are intended for export, according to the announcement. These figures describe the broader MITP business community, which includes technology companies beyond the startup sector.

New funding mechanisms target early-stage development

MITP’s newly launched Moldova Innovation Fund has an initial budget exceeding €1 million and plans four funding rounds over 24 months. It targets pre-seed and early-stage technology projects, including validation, prototyping and product development. The programme is intended to support founders before they reach the stage of raising larger commercial investment rounds.

Moldova has also introduced a fund-of-funds mechanism with a first-loss tranche, designed to use public resources to reduce investor risk and attract private and institutional capital. The mechanism provides for venture-capital and convertible investments in technology startups and innovative companies. Planned public contributions total 180 million Moldovan lei for 2026–2027, approximately €9 million, with an estimated overall capital mobilisation of 600 million lei, or about €30 million. These amounts represent planned contributions and a mobilisation estimate, rather than capital already deployed.

Organisations supporting founders include Dreamups, XY Partners, Yep! Moldova and Technovator, alongside Startup Moldova Foundation. Their programmes cover entrepreneurial education, acceleration, investor access and international market development. The Innovate Moldova programme supports these organisations and seeks to connect early-stage business development with investment and overseas market opportunities.

Startup Moldova Foundation also maintains ecosystem data, connects companies with investors and industry experts, provides legal support and works with the Moldovan professional diaspora. Its Startup Moldova Summit 2026 brought together 2,000 participants, more than 60 speakers and 150 startups in Chișinău, according to the foundation. Forty-two startups presented products in Startup Alley, while the national Startup World Cup competition offered a route to the global final in San Francisco.

The higher ranking comes as Moldova expands the financing and institutional support available to technology businesses. The next test will be whether these programmes produce sustained investment, commercial growth and completed projects. As Melniciuc noted, retaining teams and intellectual property in Moldova will also matter as companies expand internationally.

As AI infrastructure expands, the availability of the physical materials needed to build data centres and power networks is becoming an increasingly important consideration. In this written interview with Korea IT Times, Liechtenstein-based entrepreneur Monty C.M. Metzger, founder of Toto Finance, argues that copper supply and financing deserve greater attention in discussions about AI’s future. He outlines how commodity tokenization could connect producers, investors and industrial buyers, while acknowledging that Toto Finance has not yet announced a copper financing or delivery.

The following Q&A examines copper supply constraints, the opportunities and risks of tokenization, and the implications for Korean industry.

1. What evidence shows copper availability is already affecting AI infrastructure projects? How much additional copper demand comes specifically from AI and data centres, compared with broader electrification?

The bigger story is not just AI. The world is entering a period of extraordinary copper demand. By some industry estimates, humanity may need as much copper over the next 20 years as it has produced over the past 10,000 years.

For South Korea, this matters enormously. Korea's economy is built on industries that depend on reliable access to power and critical materials, from semiconductors and electronics to batteries, automobiles, shipbuilding, defence technology and now AI infrastructure. A large AI data centre can require tens of tonnes of copper for every megawatt of capacity, while the energy transition and defence industries add even more demand. I call this the Copper Crunch.

The problem is not that the copper does not exist. The problem is getting it out of the ground, financed and into the hands of industrial buyers. New mines can take nearly two decades to develop, while commodity trading and financing remain concentrated, slow and dominated by incumbent structures. Much of this industry is still operating with processes built for the last century.

AI adds another layer of pressure. S&P Global expects global data-centre capacity to grow from about 100 GW in 2022 to roughly 550 GW by 2040, with copper demand from data centres rising from around 1.1 million tonnes in 2025 to 2.5 million tonnes in 2040. Overall copper demand could rise by about 50%, creating a potential supply shortfall of around 10 million tonnes.

This is where I believe the industry is ready for its digital moment. Blockchain can create a more transparent way to finance, own and trade physical commodities and connect producers more directly with industrial buyers. That is why I founded Toto Finance. We are building a digital Mine to Market network designed to make copper easier to finance, trade and ultimately deliver to the industries that need it.

2. Where are the most serious constraints—mining, refining, manufacturing or delivery—and how could they affect project costs and timelines over the next three to five years?

The deepest constraint is mining because new supply cannot be created quickly. The IEA says average copper ore grades have fallen by about 40% since 1991, while the capital intensity of brownfield expansions has risen 65% since 2020. A new copper mine can take around 17 years from discovery to production. That is completely out of sync with the speed at which AI infrastructure and electrification are being built.

In the shorter term, manufacturing is where companies often feel the pain. Refined copper still has to become wire, cable, transformers, busbars, motors and other electrical components. So a buyer can theoretically find copper on a spot market and still wait a long time for the equipment it actually needs.

This mismatch between demand moving at technology speed and new mine supply moving at mining speed is the real problem. We cannot compress geology, but we can modernise how projects are financed and how future supply is contracted, traded and delivered. That is where Toto Finance sees an opportunity.

3. How can tokenization help bring additional copper into production? What does it improve compared with conventional project financing or long-term purchase agreements, and which constraints remain?

We start with something very tangible: Copper Now. Toto Finance is launching tokenized copper—physical, refined LME Grade A copper cathodes held in approved warehouses. Each token represents direct ownership of physical copper, with provenance back to the originating mine and the ability to redeem it for the actual metal. This brings something the traditional commodity market often lacks: a transparent digital record of where the copper came from, who owns it and how it moves through the supply chain.

From there, the same infrastructure can extend to future production. A mining company with verified reserves needs capital years before the copper reaches the market. Tokenization can connect that future production more directly with capital and industrial buyers, creating a new financing route alongside banks, commodity traders and traditional offtake agreements.

I often compare the opportunity to Uber and Airbnb. Uber created a global mobility network without owning the cars. Airbnb created a global accommodation network without owning the hotels. Toto Finance wants to build a global commodity network without owning the mines. We call it Mine to Market. Blockchain provides the ownership, provenance and settlement infrastructure connecting the mine, the capital and ultimately the industrial buyer.

For South Korea, this is bigger than financial innovation. Reliable access to copper and other critical commodities is increasingly a question of economic sovereignty. The goal is not tokenization for its own sake. It is to give Korean industry a more direct, transparent and secure way to source the physical resources it needs to compete.

4. What has Toto Finance's copper business achieved so far? Please share a named project or partner, its stage of development, financing raised and any copper produced or delivered. How does copper fit into your existing commodities business?

We spent the last two years building the technology before going after the larger opportunity. Our proprietary blockchain platform has already been proven with physical assets including diamonds, gold, silver and platinum, covering tokenization, ownership, trading and physical redemption.

Earlier this year, we consolidated that technology and IP into Toto Finance, our new US company, and moved from proof of concept into a B2B commodity strategy. Copper is the first market because the need is urgent, the market is global and the potential scale is enormous.

We are now working with mining companies and commodity partners on the first copper transactions. I recently sat down with Robert Friedland, founder of Ivanhoe Mines and one of the world's most influential mining billionaires, to discuss how copper can move through a more direct, digital Mine to Market model.

We have not yet announced a copper financing or delivery. The next milestone is clear: bring the first physical copper onto the platform, prove the model at industrial scale, and then expand that infrastructure across the commodity market.

5. How do the rights and risks differ between tokens representing underground reserves, future production, and stored copper? Who independently verifies the assets, and what happens if production or delivery fails?

I see three waves of tokenization: Money, Markets and Matter. Bitcoin and stablecoins digitized money. Then financial markets started moving on-chain. Now comes Matter, where blockchain represents ownership of the physical assets the world actually needs.

The scale is enormous. Crypto today is around $3 trillion, while gold alone is roughly a $31 trillion market. Yet only around $5 billion of gold is tokenized, less than 0.02%. A recent a16z crypto analysis shows just how early this transition still is.

Matter requires a higher standard. A token has to be tied to independently verified assets and enforceable legal rights. Stored metal can be verified through title records, warehouse documentation, insurance and physical inspection. Future production requires verified production rights, technical reports and contracts that define exactly what the holder owns and when delivery is due.

We work with experienced mining companies that understand geology and production. Our role is to add the digital ownership, provenance and financing infrastructure around those assets.

This is where blockchain becomes truly useful. It can create a permanent record of origin, ownership and transfer, while the legal structure defines what happens if delivery is delayed or obligations are not met. That combination is what can bring trillions of dollars of physical commodities into the digital economy.

6. What practical steps should Korean data-centre operators, cable manufacturers and industrial buyers consider to secure supply? Where could recycling, material substitution or direct producer partnerships help?

South Korea is already an important copper hub, with LME-approved warehouse locations in Busan, Gwangyang and Incheon. But copper sitting in Korea does not mean Korean industry owns it or has guaranteed access to it. That is the strategic issue.

Korean companies should therefore secure supply earlier and further upstream. That means diversifying sources, building direct relationships with mines, locking in future production and using recycling wherever it is economically viable. Substitution can help in selected applications, but it cannot replace the volumes of primary copper needed for AI, grids, defence and advanced manufacturing.

This is where Toto Finance fits. We are building a digital commodity network that connects mines, capital and industrial buyers more directly. Blockchain can provide a transparent record of ownership, provenance and transfer, while tokenization can help Korean buyers secure exposure to physical copper and future production before supply becomes constrained.

For Korea, the opportunity is to move from being a copper logistics hub to having greater control over supply. Toto Finance wants to provide the digital infrastructure for that shift, creating a more direct route from mine to Korean industry.

I explore this broader shift in my new book, Tokenization: Capital for Humanity, which also presents my views on commodities, tokenization and the digital economy.

Read the Korean version

Series Overview

Part 1 | Why Does AI Get Things Wrong So Convincingly?
Part 2 | Why the OXMV Loop and Evidence Ledger Are Necessary
Part 3 | Mapping Knowledge for AI with Ontology
Part 4 | OpenAI to End ChatGPT’s Custom GPT Era on December 11, 2026
Part 5 | Taking Aim at ChatGPT’s Chronic “Hallucination” Problem
Part 6 | Can AI Detect Financial Cyberattacks Before They Escalate?

By Manripungchwi Jung Yeon-tae

A series of hacking and information-leak incidents across South Korea’s financial sector has prompted financial authorities to launch an emergency industry-wide security review. According to the referenced news report, Financial Services Commission Chairman Lee Eog-weon convened an emergency meeting on October 4, 2026, in response to recent cyber threats affecting the financial industry.

He noted that multiple security incidents had been discovered within a short period across major commercial banks, savings banks and capital companies, adding that the possibility of AI-assisted cyberattacks could not be ruled out. The article identified Shinhan Bank, KB Kookmin Bank, Hana Bank and BNK Busan Bank in the banking sector, as well as Hyundai Capital, Yegaram Savings Bank and Welcome Savings Bank in the non-bank sector, as institutions where information leaks had been reported.

The fact that financial authorities moved the emergency meeting forward from its originally planned schedule suggests that the situation was being treated not merely as a series of isolated security incidents, but as a broader risk requiring industry-wide attention. What, then, can be learned from these incidents when analyzed through the framework of Financial Crime Prevention, Detection & Response AI GPT V4.0, which I developed?

Problem 1 — Attackers Did Not Focus Only on Core Financial Systems

One of the most important points in the article is not simply the reference to possible AI-assisted hacking. Financial authorities highlighted another significant characteristic of the attacks: external webpages and servers used by loan agents and employees for operational convenience—areas that may have received comparatively less security attention—were reportedly exploited as attack targets. This has important implications for financial cybersecurity.

Core banking platforms, internet banking and mobile banking systems are generally recognized as critical infrastructure and may therefore receive substantial security resources. However, if peripheral business-support systems, employee-facing platforms, externally accessible webpages, or development and operational support environments are managed with weaker controls, attackers may seek the weakest side door rather than the strongest front door.

For this reason, AI GPT V4.0 does not define the attack surface as core banking alone. It considers the following systems part of the broader attack surface whenever they contain or can access sensitive customer information:

Core Banking + ODS + CRM + Loan Origination + Call Centers + Employee Mobile Systems + Administrative Consoles + Development/Operations Support + Vendor Systems

The recent incidents therefore raise an important question about whether peripheral systems are receiving the same level of risk-based attention as core financial infrastructure.

Problem 2 — Individual Alerts May Fail to Reveal the Full Attack

Financial institutions already operate a variety of cybersecurity systems. The problem is that if signals from those systems remain isolated, an institution may fail to recognize the broader attack sequence.

For example, a login from an unusual IP address does not automatically prove malicious activity. An employee working late at night may be legitimate, while accessing customer information may also be necessary for normal business operations. But the meaning changes when activities such as the following occur within a short period:

Abnormal Login
→ New Device or Session
→ Privilege Use or Escalation
→ Customer Data Access
→ Mass Query
→ Export
→ File Creation
→ External Connection

AI GPT V4.0 is designed to correlate such activities across Account + Credential + Session + Device + IP + Dataset + Time Window rather than treating them solely as independent alerts. Its state model is:

Signal → Event → Alert → Correlated Alert → Case → Incident

An individual alert therefore does not automatically become an incident. Instead, multiple signals can be correlated and progressively escalated when the available evidence and risk context justify further investigation.

The article alone does not establish whether such sequences occurred in the recent incidents. However, when similar cyber incidents appear across multiple financial institutions within a short period, institutions should examine not only the number of security alerts generated, but also whether weak signals distributed across multiple systems could have been connected earlier.

Problem 3 — Access Must Be Distinguished from Actual Data Leakage

One of the most important principles in privacy incident analysis is:

Access ≠ Leak

A system intrusion does not necessarily mean that all customer information was exfiltrated. Viewing customer information is different from acquiring it in bulk, while an export is also different from confirmed external transmission. AI GPT V4.0 therefore structures a privacy incident through the following stages:

Exposure → Initial Access → Privilege/Discovery → Data Access → Mass Acquisition → Collection → Staging → Exfiltration → Disclosure/Abuse

An investigation should therefore go beyond a simple statement such as “the system was hacked” or “information was leaked.” It should determine which system was accessed, which dataset was queried, how much information was acquired, whether an export occurred, whether a file was created, and whether the information was actually transmitted outside the organization. Each of these should be established through evidence.

According to the article, the Financial Services Commission stated that, at that point, there was no confirmed indication that highly sensitive information directly usable for unauthorized payment transactions had been leaked. The potential scope of harm should therefore not be expanded beyond what the available evidence supports.

Problem 4 — The Risk Does Not End with the Initial Information Leak

The article also highlighted another important concern. Even if the leaked information cannot immediately be used for unauthorized payment transactions, authorities warned that it could potentially be exploited for secondary crimes such as voice phishing or smishing. This is an important element of the V4.0 framework.

A personal-data breach does not necessarily end when the privacy team completes its incident response. The leaked information may become an input for subsequent financial crime. A potential risk chain may look like this:

Customer Data Leak
→ Identity Theft
→ Credential Theft
→ Account Takeover
→ Unauthorized Transfer / Fraudulent Loan
→ Mule Account
→ Money Laundering

Financial institutions may therefore need to monitor potentially affected customers for subsequent anomalies such as unusual password resets, contact-information changes, new device registrations, new beneficiaries, suspicious lending activity, large transfers, or rapid movement of funds. Where appropriate, such activity can be correlated across FDS, Fraud and AML monitoring.

Problem 5 — Criminals Do Not Follow a Financial Institution’s Organizational Chart

Inside a financial institution, cybersecurity may be handled by the security team, privacy by the data-protection function, suspicious transactions by FDS, money laundering by AML, and internal controls by audit or compliance. Attackers do not follow these organizational boundaries.

A stolen credential can lead to customer-data access, while leaked information can facilitate identity theft and account takeover. That may be followed by fraudulent lending or unauthorized transfers, after which funds may move through mule accounts and potentially into money-laundering channels.

AI GPT V4.0 therefore integrates:

Cyber + Privacy + FDS + Fraud + AML + Internal Control + Forensics

within a unified incident model. In situations where similar cyber threats appear across multiple financial institutions, department-by-department analysis may not be sufficient. A stronger Incident Correlation framework is needed to connect alerts and evidence generated by different functions into a single Case or Incident where appropriate.

Problem 6 — Why AI-Based Defense Matters in the Age of AI-Assisted Attacks

The article reported that the Financial Services Commission could not rule out the possibility that AI had been used in the attacks. This point deserves attention. However, under the V4.0 framework, this does not mean that AI use in every incident has been conclusively established. It remains a claim requiring verification.

Nevertheless, if attackers increasingly use AI to discover vulnerabilities more rapidly, automate reconnaissance, or expand the scale of attacks, defenders may also find it increasingly difficult to rely solely on people manually reviewing enormous volumes of logs and alerts. The purpose of defensive AI should not simply be to generate more alerts.

Its more valuable role is to:

Identify which signals are connected,
distinguish normal business activity from anomalous behavior,
detect dangerous sequences earlier, and
prioritize the Cases that require human investigation.

Problem 7 — After an Incident, Ask Why the Existing Detection System Failed

When an incident occurs, investigating the attacker’s techniques is not enough. The financial institution should also determine why its existing detection framework failed to identify the threat in time. AI GPT V4.0 examines potential False Negative causes such as:

Logging Gap — The necessary logs were not available.
Coverage Gap — The affected business-support system was outside the monitoring perimeter.
Rule Gap — No suitable detection rule existed.
Correlation Gap — Multiple alerts existed but were never connected.
Vendor Blind Spot — Third-party or peripheral systems were insufficiently monitored.
Alert Closure Failure — An alert was generated but closed without adequate supporting evidence.

The referenced article does not provide enough evidence to determine which, if any, of these failures occurred at the individual institutions. They should therefore be treated as investigative questions rather than confirmed findings.

But the key question after an incident is clear: not only “Why did we fail to prevent the intrusion?” but also “Why did we fail to detect it earlier as it progressed toward customer-data access and broader harm?”

Problem 8 — A Golden Hour Response Framework Is Essential

If an attack is still active and there is evidence of ongoing external data transmission, privileged-account compromise, major fund outflows, or possible destruction of evidence, an institution cannot always wait for the investigation to reach a final conclusion. AI GPT V4.0 applies the following response sequence:

CONTAIN → PRESERVE → IDENTIFY → PROTECT → ESCALATE → VERIFY LEGAL DUTIES → INVESTIGATE

The objective is to limit further damage while preserving original logs and evidence, identifying affected systems and customers, and reviewing applicable legal reporting and notification obligations. A crucial principle is:

Risk ≠ Evidence Strength ≠ Urgency

Evidence may still be incomplete while the urgency of the situation is CRITICAL. Incomplete evidence should not automatically justify delaying containment when active harm is occurring.

Could AI GPT V4.0 Have Prevented These Incidents?

This question requires caution. It would not be appropriate to claim, “If AI GPT V4.0 had been deployed, all of these financial-sector incidents would have been prevented.” No AI system can guarantee the prevention of every cyberattack.

Actual detection effectiveness depends on the availability of logs, integration with IAM/PAM, DB/DAM, DLP, EDR and network telemetry, the monitoring coverage of peripheral systems and vendors, data quality, and the institution’s operational response capabilities.

A better question is: “Not whether every intrusion could have been prevented, but whether the institution could have received an earlier warning before the intrusion developed into a major information leak or financial loss.”

If AI GPT V4.0 had been sufficiently integrated with IAM/PAM, DB/DAM, DLP, EDR, network, HR/vendor data and peripheral business systems, it could potentially have supported earlier warning, Correlated Alert generation, and Case/Incident escalation at certain stages involving abnormal logins, suspicious credential use, customer-data access, mass queries, exports, file creation, or external transmission.

However, determining the earliest point at which each real incident could have been detected would require analysis of the original logs from the affected institutions.

The Most Important Lesson from the Recent Incidents

From a V4.0 perspective, the most important lesson from the recent financial-sector incidents is not simply that “AI-assisted hacking may have arrived.” The more fundamental issue is whether every system that stores or can access customer information is being managed at an appropriate security level.

Even if the core banking environment is highly secure, a comparatively weak employee support system, external webpage, loan-agent platform, vendor credential, or other peripheral connection may become a vulnerability affecting the entire security architecture. Financial institutions therefore need to evolve:

from Core-System-Centric Defense
to Enterprise-Wide Attack Surface Management,

from Individual Alert Detection
to Cross-System Incident Correlation,

and from Post-Incident Investigation
to Continuous Monitoring and Early Warning.

This is also the direction envisioned by Financial Crime Prevention, Detection & Response AI GPT V4.0:

Prevention → Continuous Monitoring → Early Warning → Incident Correlation → Golden Hour → Forensics → Fraud/AML → Root Cause → Remediation → Validation

The objective is to move beyond determining what went wrong after an incident and instead identify and connect the small warning signals that may already be emerging before a major incident occurs.

The financial sector does not need an AI system that automatically declares who is a criminal. It needs AI that can connect fragmented risk signals, explain why they matter through evidence, and help people make faster and more accurate decisions before those signals develop into major financial harm.

The recent financial-sector cyber incidents demonstrate why such a capability is becoming increasingly important.

Analytical Note

This article applies the analytical framework of Financial Crime Prevention, Detection & Response AI GPT V4.0 to the facts presented in the referenced news article. It does not treat specific internal-control failures, detailed attack sequences, the use of AI in individual attacks, or the identity of threat actors as confirmed facts where the source article does not establish them.

The analysis follows the principles:

Red Flag ≠ Confirmed Crime
Alert ≠ Case
Case ≠ Incident
Access ≠ Leak
Account User ≠ Actual Actor
Risk ≠ Evidence

The purpose is to distinguish reported facts from analytical hypotheses and to identify areas requiring further evidence and human review.

About the Author

Manripungchwi Jung Yeon-tae is a professional-knowledge AI planner, designer, and GPT development specialist. He currently serves as President of the Construction Industry Information Institute and develops professional GPTs across construction, law, taxation, finance, management, medicine, science, history, and philosophy. His research focuses on verified professional-knowledge AI combining ontology, XAI, the OXMV Loop, and the Evidence Ledger.

Read the Korean version

By Im Byung-min (IBM), Columnist

Artificial intelligence (AI) is changing the foundations of corporate competitiveness. Its applications are expanding from document drafting and information retrieval to strategy, operations, customer experience, and product and service development. Companies therefore face a broader task: integrating AI into how they organize and run their businesses. The ability to create new value with AI could become a significant source of competitive advantage.

Drafting documents, summarizing meetings, and deploying chatbots for customer inquiries are useful starting points. These applications reduce repetitive work and free up employees’ time. Yet competitors can adopt similar tools, making lasting differentiation difficult to achieve through adoption alone. The next challenge is to turn those efficiency gains into capabilities unique to the business.

Connecting Proprietary Data and Operational Knowledge

Capital, facilities, technical expertise, talent, and brands remain important foundations of competitiveness. The ability to combine a company’s accumulated data with AI adds another dimension. Customer purchasing patterns, production and quality data, research and development records, sales expertise, and supply chain information reflect years of operational experience. Lessons from successes and failures, together with employees’ practical knowledge, are also assets that competitors may struggle to replicate quickly.

Whether these assets are available in a form the entire organization can use is a separate question. Data may be scattered across departmental systems, while the reasoning behind important decisions remains in individuals’ memories. Under these conditions, adopting AI may leave much of the company’s knowledge untapped. An AI strategy should therefore address model selection alongside an assessment of what the company knows, what data it holds, and where its strengths lie.

Data’s value depends on its quality and connectivity as well as its volume. Companies need to identify the information they already hold, fill gaps relevant to their work, and connect data across systems. Practical experience must also be documented and organized so that colleagues and AI systems can draw on it. When this knowledge becomes part of everyday decision-making, proprietary data can contribute to a stronger competitive position.

A salesperson using AI to draft a report improves the efficiency of an individual task. A company can go further by linking customer transaction history and market analysis to the identification of potential needs, product recommendations, and proposal preparation. Extending this process to follow-up activities broadens AI’s role across the sales cycle. Workflow redesign connects the information and time gained at each stage to subsequent decisions and actions.

AI agents can retrieve relevant information and perform analyses and tasks across multiple systems to pursue a defined goal. Companies can use these capabilities to build workflows in which people and AI work together. They must decide which tasks to automate and where employees should focus their attention. Reducing time spent on routine information processing and reporting can free up capacity for customer relationships, creative problem-solving, strategic judgment, and identifying business opportunities.

Turning Small Experiments into Customer Value and Business Results

This transformation requires coordination across the business. It should be treated as a company-wide management priority involving both executives and frontline employees. Production, sales, research and development, and other teams need to identify tasks that AI could improve, while management connects those initiatives to corporate strategy. A structure for documenting and sharing implementation experience helps turn departmental lessons into organizational knowledge.

Starting with a clearly defined task is more practical than attempting to build a comprehensive system from the outset. Small experiments with measurable results help establish which approaches work. When outcomes fall short of expectations, companies should examine the causes and improve the underlying data and procedures. Proven applications can then be extended to other departments and workflows, allowing the approach to develop around the company’s needs.

Performance evaluation should examine the benefits customers receive alongside cost savings and productivity gains. Relevant questions include whether AI improves understanding of customer needs, enables more personalized services, or creates opportunities for product and service development. Directing resources freed up by greater efficiency toward these improvements can help AI investment support business growth. The same process can reveal new revenue models grounded in the company’s data and experience.

Companies using the same AI technology may achieve different results. Some may focus on productivity in individual tasks, while others extend its use across products, services, and operations. Outcomes depend in part on how effectively data is used, how practical knowledge is captured, how responsibilities are assigned to people and AI, and how lessons from implementation are shared. Developing these capabilities deserves attention alongside choosing the technology.

Management’s questions must also broaden. Executives should consider which AI tools to adopt, what distinctive value their company offers customers, and how the technology can strengthen that value. They should explore which products and services could emerge from combining existing data, expertise, and experience. Opportunities that competitors might pursue first also merit attention, with priority given to concrete initiatives that build on the company’s strengths.

Future competitiveness depends on how effectively a company brings its data, technology, experience, people, and culture together in applying AI. It must decide what to improve and what new value to create. Experimentation and validation can then guide the expansion of successful applications. The ability to redesign workflows and business operations with AI will help build a foundation for sustained competitive advantage.

Read the Korean version

The Korean telecom operator said on Oct. 6 that it will work with Bango to introduce a wider range of global AI and content services that have previously been difficult for domestic users to access.

Youdok is LG Uplus’ subscription platform that allows customers to subscribe to and manage services across AI, over-the-top streaming, education and lifestyle categories in one place.

As the first result of the partnership, LG Uplus plans to launch an AI-powered video editing service on Youdok in November. The company will subsequently expand its lineup to include generative AI, productivity tools, creative AI and other global content services.

AI Becomes a New Growth Area for Subscription Services

The partnership comes as consumer demand for subscription services expands beyond entertainment content toward AI tools designed to improve productivity, work efficiency and creative activities.

LG Uplus has been strengthening Youdok’s AI portfolio in response to this shift. The platform already offers AI services across search, design, video editing and education, including Google AI Pro.

It also operates “Youdok Pick AI,” which allows customers to choose among multiple AI subscription services according to their needs.

Through the partnership with Bango, LG Uplus plans to add overseas AI services that are not yet widely available in South Korea or require users to subscribe and pay through separate international platforms.

The two companies will cooperate in identifying AI and content services in Korea and overseas, integrating new subscription products, developing bundled offerings and expanding the digital subscription ecosystem.

Bango’s DVM to Accelerate Service Integration

A key part of the partnership will be Bango’s Digital Vending Machine, or DVM, a subscription bundling platform that connects digital service providers with telecom operators and other distribution partners.

LG Uplus plans to use the DVM to streamline contracts, technical integration and product launches for new services on Youdok.

The company expects the system to shorten the process required to introduce global AI and content services, while allowing it to broaden the range of subscription options available to customers.

LG Uplus and Bango also plan to combine Bango’s global service network with LG Uplus’ product planning capabilities to develop subscription packages tailored to different customer needs and use cases.

The companies will initially focus on areas drawing strong consumer interest, including generative AI, productivity AI and creative AI, while gradually adding specialized global content services.

Expanding Youdok Beyond Traditional Content Subscriptions

LG Uplus has been expanding Youdok from a platform centered on conventional digital subscriptions into a broader marketplace covering AI and productivity tools.

The latest partnership is designed to give Korean consumers easier access to global services without having to separately search overseas websites or manage individual subscriptions and payment processes.

Cho Yong-sung, vice president in charge of partnership business at LG Uplus, said customer demand for subscriptions is rapidly expanding beyond content into AI and productivity services.

“Through this partnership, we will make it easier and more convenient for Korean customers to access a diverse range of global AI services while continuing to strengthen Youdok’s competitiveness in AI subscriptions,” Cho said.

Paul Larbey, CEO of Bango, said the company will support LG Uplus in rapidly adding global services through the DVM and creating subscription packages tailored to customer demand.

Rather than positioning jari simply as another imported liquor brand, CJ CheilJedang is using Korean cuisine, fine dining and cocktail culture as entry points for introducing the brand to U.S. consumers.

The company held a jari launch event on Sept. 29 at Niku X, a restaurant in downtown Los Angeles. More than 100 representatives from local restaurants, bars and the broader food and beverage industry attended the event, where they sampled three jari products and explored pairings with Korean food.

The event also highlighted jari’s collaborations with small and independent Korean distilleries. By bringing those producers into its overseas expansion strategy, CJ CheilJedang is positioning jari as a platform for showcasing a broader range of Korean spirits rather than promoting a single brand in isolation.

A representative from Hojokban LA, a Korean fusion restaurant operated by GFFG, said the company’s approach to working with distilleries stood out, adding that the partnership model could help communicate the authenticity and value of Korean spirits to American consumers.

The Los Angeles event followed a New York gathering held on Sept. 23 at Gui Steakhouse in Manhattan. The event brought together representatives from prominent Korean restaurants and bars, including Jungsik, Joo Ok, Jua, Mari and Take 31.

Korean spirits were presented alongside food and cocktail pairings during the New York event. Chef Sungcheol Shim, who operates Michelin-starred restaurants Mari and Kochi, and bartender Hong Doo, a mixologist from jari brand ambassador Pine&Co, introduced different ways of pairing Korean distilled spirits with contemporary Korean cuisine and drinks.

Interest generated by the New York launch has already translated into new restaurant and bar placements. Jari is set to be available at a number of prominent New York dining venues, including Michelin-starred restaurants such as Jungsik and Joo Ok, as well as C as in Charlie, Take 31, Seoul Salon and Nari.

The company plans to expand consumer access to the brand from October, working with local partners to gradually broaden its retail and foodservice presence.

Industry observers say the strategy reflects a broader shift in the way Korean spirits are being introduced to international consumers. As Korean cuisine gains visibility in the U.S., pairing spirits with established dining experiences can provide consumers with a more accessible way to discover unfamiliar categories.

For jari, the challenge will be to turn initial interest among chefs, bartenders and Korean dining operators into sustained consumer demand. The U.S. spirits market is highly competitive, and long-term growth will depend not only on product quality but also on distribution, brand recognition and the ability to establish a distinct identity beyond the Korean food trend.

CJ CheilJedang believes the growing popularity of Korean food provides an opportunity to introduce Korean spirits as part of a broader cultural and culinary experience.

“Through our launches in New York and Los Angeles, we have seen strong interest from the local dining industry in Korean spirits and their potential for growth,” a CJ CheilJedang representative said. “We aim to expand jari’s role as a platform that introduces the appeal of Korean spirits to global consumers while creating new dining experiences that pair them with Korean food.”

With New York and Los Angeles now serving as key launch markets, CJ CheilJedang plans to expand jari’s presence through partnerships with restaurants, bars and other local food and beverage businesses, while continuing to work with Korean distilleries to bring a wider range of K-Spirits to the U.S. market.

By Manripungchwi Jung Yeon-tae

Following a series of cyber intrusions and data exposure incidents across banks, savings banks, and capital companies, financial authorities have launched an emergency sector-wide review. One of the most important characteristics of the recent incidents is that attackers reportedly targeted external web pages, servers, and other operational systems used by loan agents and employees—areas that may have received less security attention than core banking systems. Authorities have also raised the possibility that artificial intelligence may have been used in the attacks, reinforcing the need for financial institutions to rethink how they prepare for and respond to cyber incidents. But the first priority for a financial institution should not be to determine whether the attacker used AI.

The immediate priorities are to determine whether the incident is still active, prevent further harm, preserve evidence before it disappears, and establish exactly what data was accessed, acquired, exported, or transmitted. Applying the Cyber, Privacy, Forensics, Fraud, FDS, AML, and internal-control framework of the Financial Crime Prevention, Detection, and Response AI GPT, a financial-sector cyber incident can be handled through the following structured process.

1. The First Priority Is Not Root-Cause Analysis, but Containment

During the early stages of an incident, organizations can easily become focused on determining who the attacker was or which attack tool was used. However, if customer information is still being queried, downloaded, or transmitted—or if a privileged administrator account has been compromised—preventing additional harm must come first. The Financial Crime Response AI GPT structures this as a Golden Hour workflow:

CONTAIN → PRESERVE → IDENTIFY → PROTECT → ESCALATE → VERIFY LEGAL DUTIES → INVESTIGATE

The first step is CONTAIN. Potentially compromised accounts, sessions, administrative privileges, and external access routes should be identified, and additional customer-data access, download, or external transfer should be restricted where appropriate. This does not mean shutting down every system indiscriminately. Broad shutdowns can cause unnecessary service disruption or even destroy evidence. Containment should therefore be performed under the authority of the institution’s CISO, information-security team, IT operations, compliance, legal, audit, and other responsible functions.

The second step is PRESERVE. Potentially relevant evidence should be secured as quickly as possible, including: IAM, MFA, and PAM logs; database query and export records; EDR telemetry; file and USB activity; firewall, proxy, VPN, and DNS logs; application logs; DLP records; email and cloud activity; and, where relevant, physical access records. If logs are overwritten through normal retention cycles or systems are altered after discovery, it may become significantly more difficult to determine the true scope of the incident. Forensic practice should distinguish among Original Evidence, Forensic Copies, Working Copies, and Derived Evidence, with appropriate hash values and chain-of-custody records maintained.

2. “Access” and “Leak” Must Be Treated as Different Facts

One of the most common errors in privacy incidents is to assume that a large-scale query automatically means that all queried information was exfiltrated. Access to customer information and external exfiltration are not the same event. The Financial Crime Response AI GPT analyzes privacy incidents through a Privacy Kill Chain:

P0 Exposure → P1 Initial Access → P2 Privilege/Discovery → P3 Data Access → P4 Mass Acquisition → P5 Collection → P6 Staging → P7 Exfiltration → P8 Disclosure/Abuse

For example, even if logs show that an attacker queried 300,000 customer records, that fact alone does not prove that 300,000 records were exfiltrated. Investigators must determine, step by step, whether there was: a database query; an export; file creation; compression or encryption; staging in an internal or temporary location; and actual transmission to an external IP address, cloud service, email account, USB device, or other destination. An incident report should therefore distinguish at least three separate scopes: Confirmed Access Scope; Confirmed Download or Export Scope; Confirmed External Transfer Scope. These may not be identical.

3. Looking Only at Core Banking Systems Can Leave Major Blind Spots

Another major lesson from this incident is that attackers do not need to compromise a core banking platform to create a serious financial-sector incident. Customer information may also reside in, or be accessible through: loan-agent systems; CRM platforms; ODS environments; employee web systems; call-center platforms; administrator consoles; development and operations support environments; and third-party or vendor systems. Under the Financial Crime Response AI GPT framework, these Core and Non-Core systems are treated as part of the same overall attack surface. Financial institutions should therefore be able to answer questions such as:

The security model therefore needs to move beyond a narrow “protect the core system” mindset toward an Attack Surface Inventory that includes every system capable of storing, processing, or accessing customer information.

4. A Single IP Address Is Not Enough to Identify the Attacker

IP addresses, user accounts, and devices are important indicators during a cyber investigation. But activity performed through a particular account does not necessarily prove that the account owner was the actual actor. Possible alternative explanations include: credential theft; shared accounts; credential sharing; VPN or proxy use; automated processes; or compromised devices. The actual actor therefore needs to be investigated through a broader chain such as:

Credential → Session → Device → IP → Privilege → System → Dataset → Query → Export

If a privileged account was involved, investigators should determine when the privilege was granted, who granted it, which device and IP address were used, and whether the supposed user was actually present or active at the relevant time.

Account User ≠ Actual Actor.

5. Claims of an “AI-Powered Attack” Must Also Be Evidence-Based

The possibility that AI was used in recent financial-sector attacks has received significant attention. But there is an important difference between saying: “AI-assisted attack methods cannot be ruled out” and saying: “AI conducted the attack.” The latter requires supporting technical evidence. Potential indicators may include: attack speed and repetition; automated payload generation patterns; high-volume endpoint discovery; API-call behavior; tool execution records; automation frameworks; and other traces consistent with AI-assisted or machine-driven attack activity.

The Financial Crime Response AI GPT classifies such claims using evidence-based fact states:

Claims involving AI should be expressed with the same discipline.

6. After the Cyber Incident Ends, Secondary Financial Crime May Begin

A privacy incident does not necessarily end when systems are restored. Exposed information may later be used in: voice phishing; smishing; account takeover; identity theft; fraudulent loans; card fraud; or other forms of financial crime. The Financial Crime Response AI GPT models this post-leak risk as:

Data Leak → Identity Theft → Synthetic Identity → Account Takeover → Unauthorized Transfer → Fraudulent Loan / Card Fraud → Mule Account → Money Laundering

For that reason, FDS and AML monitoring should be strengthened after a major data exposure incident. For example, if a customer account shows the following sequence: Contact Information Change → Password Reset → New Device Registration → Transfer Limit Increase → New Beneficiary Registration → High-Value Transfer, these events should not be reviewed as isolated alerts. They may represent a single Account Takeover case. Failure to correlate those events can lead to missed fraud.

7. Institutions Must Ask: “Why Did We Fail to Act on Earlier Alerts?”

A cyber investigation should not focus only on what the attacker did. It should also examine whether internal warning signs had already appeared but were ignored, prematurely closed, or repeatedly overridden. Under the Financial Crime Response AI GPT framework, Alert Closure and Override are themselves monitored as potential control and investigation signals. Examples include: the same administrator repeatedly closing high-risk alerts; alerts being closed unusually quickly; high-risk alerts being dismissed without supporting evidence; repeated alerts involving the same account not being aggregated into a Case; and post-incident review showing that warning signs had existed for weeks or months.

In such situations, the issue may not be only a cybersecurity failure. It may also represent a failure in detection, monitoring, escalation, or governance.

8. Vendors and Development Accounts Must Be Investigated at the Same Level

The financial institution’s attack surface includes its vendors, contractors, outsourcing partners, and development providers. The following chain should therefore be examined:

Vendor → Contract → Employee → Credential → Privilege → Device → IP → System → Customer Data

Important risk indicators include: vendor accounts remaining active after contract termination; former contractor accounts not being revoked; developers accessing production databases directly; shared credentials; privileged accounts without MFA; and excessive long-term access rights. Third-party risk cannot be separated from internal cybersecurity risk simply because the user works outside the financial institution.

9. Root Cause Analysis Must Go Beyond “More Security Training”

After the immediate investigation, institutions need to determine the root causes of the incident. The Financial Crime Response AI GPT separates root causes into areas such as: People; Process; Technology; Governance; Control; Data; and External factors. For example, the immediate cause of an incident may be an external attacker stealing credentials. But the deeper root causes may include: lack of MFA; excessive administrator privileges; unrevoked vendor accounts; absence of bulk-export restrictions; gaps in log collection; repeated alert overrides; or a vulnerable system left unpatched for an extended period.

An incident like this should not be closed with a generic recommendation such as “strengthen employee security training.” Corrective actions should instead be structured as:

Finding → Root Cause → Action → Owner → Due Date → Completion Evidence → Validation → Sustainability

And institutions must distinguish between simply implementing a change and proving that the change works.

IMPLEMENTED ≠ VALIDATED

A 24-Hour Incident Response Framework for Financial Institutions

During the first 24 hours after detection, the following sequence can be used as a practical operating framework.

0–1 Hour

Identify suspected compromised accounts and sessions. Determine whether active exfiltration is occurring. Check for privileged-account compromise. Activate the incident-response team.

Restrict affected access and external transfer paths where necessary. Prevent log rotation or automatic deletion.

1–4 Hours

Secure IAM, MFA, and PAM logs. Preserve database query and export logs. Preserve EDR, network, DLP, cloud, and email evidence. Estimate the preliminary scope of customer-data access.

Map relationships among suspected accounts, devices, IP addresses, and systems. Begin building the incident Master Timeline.

4–8 Hours

Differentiate Access, Query, Export, Staging, and Exfiltration. Identify affected datasets and estimate the scope of impacted data subjects. Determine whether the same IOC, credential, device, or session appears in other systems. Expand the review to vendor and contractor accounts. Conduct an emergency threat hunt across comparable systems.

8–12 Hours

Assess the possibility of customer financial harm. Consider strengthening FDS rules related to account takeover, identity theft, voice phishing, and fraudulent lending. Prepare an initial executive report. Have legal, compliance, privacy, and information-security teams review applicable reporting and notification obligations. Consider external forensic support where necessary.

12–24 Hours

Prepare a preliminary incident-scope assessment. Update the Master Timeline. Build an Evidence Matrix. Document primary attack-path hypotheses and counter-hypotheses.

Search for additional potentially affected customers. Begin preliminary Control Failure analysis. Prepare materials for potential supervisory or regulatory engagement. Establish the formal investigation plan for the period after the first 24 hours.

Where Can a Financial Crime Response AI GPT Help?

An AI GPT should not replace the institution’s legal authority, incident commander, investigator, system administrator, or regulatory decision-maker. Its value lies in structuring complex information. When logs and transaction records are provided, it can help reconstruct timelines, correlate multiple alerts into a single case, and map relationships among accounts, devices, IP addresses, systems, datasets, and transactions. It can also support the preparation of: incident reports; Evidence Matrices; evidence inventories; investigation questionnaires; executive briefings; regulatory response materials; Root Cause Analysis reports; and remediation plans, all from a consistent factual foundation.

The objective is not for AI to declare who committed a crime. The objective is to: separate facts from inference, connect fragmented evidence, avoid presenting unverified claims as confirmed facts, and help human investigators and decision-makers respond more quickly and accurately. This series of financial-sector cyber incidents raises an important question. It is no longer enough for a financial institution to ask: “Is our core banking system secure?”

About the Author

Manripungchwi Jung Yeon-tae is a professional-knowledge AI planner, designer, and GPT development specialist. He currently serves as President of the Construction Industry Information Institute and develops professional GPTs across construction, law, taxation, finance, management, medicine, science, history, and philosophy. His research focuses on verified professional-knowledge AI combining ontology, XAI, the OXMV Loop, and the Evidence Ledger.

Read the Korean version

Part 1 | When a Company Fails to Pay for a Transaction
Part 2 | Continued Use of a Business Name and the Transferee’s Liability
Part 3 | A Company Director’s Liability for Damages
Part 4 | The Person Hiding Behind a Company
Part 5 | What to Consider After Being Assaulted
Part 6 | What Is an Appropriate Criminal Settlement Amount?
Part 7 | If You Cannot Remember What Happened After Drinking, Could It Constitute Quasi-Rape?
Part 8 | Freelancers and Severance Pay
Part 9 | What to Consider When Assigning a Claim to Secure Business Financing
Part 10 | Can You Divorce a Foreign Spouse Who Has Disappeared?
Part 11 | Can a Tenant Terminate a Lease If the Landlord Delays Repairs?
Part 12 | Can a Subcontractor Receive Payment If the Main Contractor Disappears?
Part 13 | What to Know Before Your First Police Interview
Part 14 | Can a Buyer Recover a Deposit Before Receiving Approval for a Contract in a Land Transaction Permit Zone?
Part 15 | Is Entering Another Person’s Home While Concealing a Criminal Purpose Trespassing?
Part 16 | Impersonating Another Person on Instagram Can Lead to Criminal Punishment
Part 17 | When Does Indecent Assault Occur?
Part 18 | When Repairs Are Required After a Water Leak, Can the Tenant Decide How the Work Is Done?
Part 19 | How Far Does a Real Estate Agent’s Duty to Explain Extend?
Part 20 | Does Swearing in a Residential Area Constitute Criminal Insult?
Part 21 | Can Damages Be Claimed for Incorrect Tax Advice Without a Formal Engagement Agreement?
Part 22 | Can Families Claim Damages for the Exhumation of an Ancestral Grave?
Part 23 | How Far Does Individual Liability Extend in Cambodia-Based Scam Rings?
Part 24 | When Does Filming Sexual Activity Between Partners Constitute a Crime?
Part 25 | Childcare Corporation Assets Can Now Pass to a Designated Recipient After Dissolution

By Hong Seung-kwon, Principal Attorney, PAN Law Office (skh@law-pan.com)

Once a social welfare corporation is established, property contributed by its founder no longer remains the founder’s personal property. When such a corporation is dissolved, its remaining assets must generally revert to the national or local government in accordance with its articles of incorporation. Article 27 of the Social Welfare Services Act establishes this principle. Even if a founder contributed personal land and buildings decades ago to establish a social welfare corporation and operate a childcare center, dissolution did not allow the founder to reclaim those assets.

A significant exception took effect on May 12, 2026, for certain social welfare corporations operating childcare centers. The newly enacted Article 43-3 of the Infant Care Act allows a qualifying corporation to transfer its residual assets to a designated recipient instead of having them revert to the national or local government upon dissolution. The amendment reflects the sharp decline in the number of young children caused by Korea’s low birth rate and the resulting financial difficulties facing childcare centers. It is intended to ease a problem that prevented corporations from dissolving even when continued operation had become effectively impossible because all their assets would otherwise revert to the government.

A particularly important issue is the range of people who may receive the residual assets. When applying for authorization to dissolve, the corporation must submit a residual asset disposition plan identifying the intended recipient and explaining the reason for the transfer. The “person designated in the residual asset disposition plan” may be the original contributor, a relative or even an unrelated third party. The assets do not have to be returned solely to the person who originally contributed them.

The property subject to transfer is limited to the residual assets remaining after liquidation, rather than all assets held by the corporation. During liquidation, the corporation must first repay all its debts. If any property was acquired with subsidies from the national or local government, the corporation must return the required subsidy amount after accounting for depreciation in accordance with applicable laws. Only the property remaining after the debts and subsidy obligations have been settled may be transferred to the designated recipient.

Separate tax obligations may arise during the transfer. Depending on the recipient and the nature of the property, the transfer may trigger gift tax, capital gains tax or acquisition tax. The parties must therefore review the applicable tax laws and complete the required filings and payments.

Not every childcare center operator may use this special provision. The operator must be a social welfare corporation established under the Social Welfare Services Act, and its stated purpose in the articles of incorporation must be limited to the establishment and operation of childcare centers. A corporation that operates a childcare center alongside a senior welfare facility or another type of social welfare facility is generally excluded.

The childcare center operated by the corporation must also meet one of the legally prescribed conditions indicating operational difficulty. Article 37-2 of the Enforcement Rule of the Infant Care Act applies when the childcare center has already closed or suspended operations, when its average monthly enrollment rate remained below 20% during the 24 months immediately preceding the application for dissolution, or when the competent mayor or provincial governor publicly recognizes comparable circumstances that make it difficult to continue the corporation’s stated purpose.

This special provision is temporary. The addendum to the Infant Care Act limits the application of Article 43-3 to seven years from its effective date, meaning that the provision is currently scheduled to remain in force until May 11, 2033. A social welfare corporation established long ago with privately contributed assets and now struggling with declining enrollment should therefore consider whether it qualifies for the dissolution exception. That assessment should cover its articles of incorporation, the assets originally contributed, its subsidy history, enrollment rates over the preceding 24 months, existing claims and liabilities, and potential tax obligations.

About the Author

Hong Seung-kwon is the principal attorney of PAN Law Office. His practice covers civil and criminal litigation as well as emerging legal issues in the digital era. He continues to examine how technological innovation, property rights and changes in regulatory policy affect law and market order.

Read the Korean version

NVIDIA is expanding the boundaries of AI infrastructure with a new security architecture designed to monitor and control the actions of AI agents across both software and hardware. As the industry assessed the technology on October 3, a new security challenge was coming into focus: AI safety is no longer only about what a model says, but increasingly about what an autonomous agent is allowed to do when it can open files, execute code and connect to external systems.

NVIDIA unveiled its Open Agent Safety Platform on September 28, introducing a dual-layer approach intended to prevent AI agents from acting beyond defined permissions. At the center of the architecture are OpenShell, an open-source secure runtime for AI agents, and Sentry, an independent hardware-based monitoring system.

OpenShell runs AI agents inside isolated environments and controls their access to files, networks, data, application programming interfaces and other services according to predefined policies. It follows a zero-trust approach in which access is denied by default and only necessary permissions are granted. Crucially, security policies are enforced outside the agent process itself, making them harder for an agent to bypass through its own reasoning or prompts.

Sentry provides a second line of defense. Running on NVIDIA’s BlueField-4 data processing unit, it independently monitors agent activity and is designed to isolate and stop an agent within milliseconds when it attempts to cross established boundaries. NVIDIA describes the architecture as a full-stack safety system spanning models, software, processors, networks and data access.

Why AI Agents Need a New Safety Layer

The fundamental difference between generative AI and AI agents is their ability to act. When a chatbot produces an incorrect answer, a person can review it and decide not to use it. But when an autonomous agent can send emails, modify files, execute code and connect to external services inside an enterprise environment, a bad decision can become a real-world action.

Recent security incidents in the AI industry have brought that risk into sharper focus. Reuters reported that leading AI companies have been examining cases in which agents accessed external systems in unexpected ways or circumvented security controls. NVIDIA has argued that its new platform could have prevented agent activity associated with an intrusion involving Hugging Face earlier this year.

Justin Boitano, NVIDIA’s vice president and general manager of enterprise computing, told Reuters that based on what is now known about the incident, the new security platform could have prevented the breach had it been used during the initial model evaluation process.

NVIDIA CEO Jensen Huang also emphasized the broader challenge when announcing the platform. “AI’s enormous potential can only be realized when we solve AI safety,” Huang said, adding that safety and security require “full-stack engineering.”

NVIDIA’s approach is to avoid relying solely on an AI agent to follow its own safety rules. Instead, it places safeguards outside the agent’s control, allowing the system itself to determine and enforce the boundaries within which an agent can operate.

Expert Says Security and Functionality Need Real-World Testing

Restricting AI agents through software and hardware does not, however, resolve every aspect of AI safety. Somesh Jha, a professor of computer science at the University of Wisconsin–Madison, has pointed to a fundamental trade-off in NVIDIA’s approach: security controls that are too restrictive could also prevent agents from performing legitimate and useful tasks.

Speaking to the Associated Press, Jha said whether such systems can strike the right balance between security and functionality is something that “only case studies can answer.” The practical test will therefore be whether agents can continue to perform necessary work in enterprise environments while dangerous actions are reliably blocked.

That tension lies at the heart of agent security. Give an AI agent too little authority and much of the value of autonomy disappears. Give it too much and the security risk increases. The key question is therefore not simply how intelligent an agent is, but what data and tools it can access, when human approval is required and whether it can be stopped immediately when something goes wrong.

NVIDIA Moves Toward an ‘AI Safety Infrastructure’ Layer

Another important signal from the announcement is the continued expansion of NVIDIA’s business beyond computing hardware. The company has already moved beyond graphics processing units into central processors, networking, AI software and enterprise AI platforms. The Open Agent Safety Platform adds another potential layer: infrastructure for controlling the permissions and behavior of autonomous agents.

According to NVIDIA, more than 100 companies and organizations are participating in the development or adoption of technologies around the platform, including Anthropic, Microsoft, Cisco, CrowdStrike, JPMorganChase, Palantir, Palo Alto Networks, Salesforce, SAP, ServiceNow and Scale AI. OpenShell is being released as open-source software and is designed to extend beyond NVIDIA hardware to computing environments based on Arm and Intel architectures.

If enterprises eventually operate hundreds or thousands of AI agents, the question of who controls their permissions, records their actions and shuts them down when necessary could become as important as who develops or deploys the agents themselves. As the number of autonomous agents grows, agent security could emerge as an infrastructure market in its own right.

NVIDIA is therefore moving beyond providing the computing power behind AI. It is seeking a role in defining and managing the boundaries within which AI agents can act. As autonomous agents spread across enterprise systems, control of this security layer could become another important front in the AI platform competition.

In the Agent Era, What AI Can Do Matters More

The signal from NVIDIA’s announcement is clear. In the generative AI era, much of the focus was on what a model knew and what kind of answers it produced. In the agent era, security increasingly depends on which files an AI can open, which systems it can enter and which tools it is authorized to execute.

As AI becomes more capable, security cannot depend solely on the model making the “right” decision. Boundaries need to exist outside the agent itself, its actions need to be recorded, and humans must be able to intervene when necessary.

If the GPU race determined how quickly AI could compute, the next phase of competition in the agent era may increasingly revolve around how much authority AI should be given — and who controls that authority.

What NVIDIA has introduced is therefore more than another cybersecurity product. As AI moves from a technology that talks to one that acts, the company is positioning itself for an emerging infrastructure market built around controlling the boundaries of those actions.

Read the Korean version

On October 3, financial authorities in Seoul stepped up their response to a series of cyber intrusions and personal data leaks that have spread from South Korea’s major banks to savings banks and consumer finance companies. But the more important question is not how many financial institutions have been affected. It is where the attackers chose to strike.

Incidents disclosed so far point not to core transaction systems such as internet or mobile banking, but to peripheral systems connected to the internet, including employee support platforms, loan-recruitment services and third-party solutions. Rather than directly confronting the most heavily protected parts of the banking infrastructure, attackers appear to have sought weaker points around the edges.

Financial authorities are moving quickly. On October 4, the Financial Services Commission (FSC) will convene an emergency meeting at the Government Complex Seoul with representatives from across the financial sector, including banking, securities, insurance, credit finance, savings banks, mutual finance, virtual assets and fintech, as well as executives from financial institutions affected by security incidents. The FSC had initially planned to receive the results of financial companies’ internal security inspections on October 7, but brought the meeting forward after additional incidents were identified at a savings bank and a capital company.

The Government Complex Seoul in Jongno-gu, Seoul, which houses the Financial Services Commission. Financial authorities are conducting an industry-wide security review following a series of cyber intrusions targeting South Korean financial institutions. / Courtesy of Financial Services Commission

The scale of confirmed exposure is already significant. Shinhan Bank reported a leak involving information on about 25,000 customers, while KB Kookmin Bank identified exposure involving 119 customers and Hana Bank 89 customers. At BNK Busan Bank, no customer data leak has been confirmed, although personal information belonging to 11 outsourced developers was found to have been exposed.

Woori Bank and NH NongHyup Bank also detected external intrusion attempts, but blocked them, with no data leak confirmed so far. Beyond the major banks, Yegaram Savings Bank identified a potential leak involving information on an estimated 40,000 customers, while Hyundai Capital confirmed the exposure of some personal information belonging to 146 mortgage loan recruiters.

There is still no evidence that all of these incidents were carried out by a single threat actor. Investigators have yet to determine whether one group targeted multiple financial institutions or whether separate incidents surfaced at roughly the same time as the industry intensified its security inspections.

Attackers Found a Route Outside the Core Banking Network

The most important feature of these incidents is the type of systems in which vulnerabilities and exposures have been identified.

For years, financial institutions have invested heavily in protecting core transaction systems such as internet and mobile banking through network separation, strong authentication and access controls. Those defenses remain essential. So far, there has been no confirmation that the core financial transaction systems themselves were directly compromised in the latest incidents.

The problem lies beyond that core.

As financial services have become increasingly digital, the number of connections surrounding core banking infrastructure has continued to grow. Employee mobile platforms, loan-recruiter portals, third-party software, application programming interfaces (APIs), and systems operated by contractors and business partners can all provide routes to customer information or internal data.

Attackers do not need to confront the most heavily defended system. If they can reach sensitive information through a less protected application or service, that becomes the easier target.

That is why the FSC, at an emergency response meeting on October 2, instructed financial institutions to identify all IT assets and services accessible from outside their networks. Companies were told to review authentication and access controls regardless of whether a system was designed for customers or internal operations, and to minimize unnecessary exposure of information.

Authorities also called for rapid sharing of malicious IP addresses, attack techniques and attempted intrusion data among financial companies and relevant agencies. The move reflects a growing recognition that individual institutions acting alone may struggle to respond quickly enough to rapidly evolving cyber threats.

The Bigger Issue Is Attack Automation, Not the ‘AI Hacking’ Label

The possibility that artificial intelligence may have played a role in the attacks has also drawn attention.

A server believed to be associated with the attack on Shinhan Bank reportedly contained traces linked to a Chinese-language AI-based autonomous penetration-testing environment. Security industry experts have raised the possibility of a connection to an open-source autonomous penetration-testing tool known as ARTEX AI. Such tools can use large language models and multi-agent technologies to automate tasks including reconnaissance, vulnerability discovery, attack-path planning and security-tool execution.

But evidence of such an environment is not the same as proof that AI was actually used to carry out the attack. It has not been confirmed that ARTEX AI was deployed in the Shinhan incident, nor has it been established that the attacks against multiple financial institutions came from the same threat actor.

It is therefore premature to describe the incidents as an “AI attack on the financial sector.”

The more important issue is what AI could enable attackers to do next.

Systems and vulnerabilities that previously required significant manual effort to discover could increasingly be scanned by AI agents. If those agents can also help map attack paths and automate parts of the exploitation process, the time and resources needed to mount attacks could fall substantially.

A financial institution has to defend thousands of systems. An attacker needs to find only one weak point. If AI accelerates that search, financial cybersecurity will have to detect and close vulnerabilities much faster than before.

Financial Security Must Expand Beyond the Core Network

The incidents raise a fundamental question about what financial institutions should consider a critical security asset.

Strong protection of core banking networks remains indispensable. But a secure core network no longer means that the entire institution is secure. If an employee application, an old lookup page or a third-party solution is connected to customer information or internal data, it too becomes an entry point from an attacker’s perspective.

The FSC’s demand that financial institutions identify all externally exposed IT assets reflects this changing reality. Financial firms will increasingly need to manage their entire digital attack surface through measures such as Attack Surface Management (ASM), multi-factor authentication, least-privilege access, detection of abnormal bulk queries and stronger API controls. Contractors, business partners and third-party solutions also need to be brought within the same security framework.

The latest incidents are forcing South Korea’s financial sector to reconsider what financial security actually means. Banks do not only have to protect their core networks. A single weak employee system or internet-connected loan portal can be enough for sensitive information to escape.

Attackers do not fight the strongest door. They look for the weakest one. The task for financial institutions now is to find and close that weakness before the attackers do.

Read the Korean version

The Consulate General of the Republic of Korea in Vancouver hosted the 2026 Korean National Day Reception on October 1 at 6 p.m. local time in the Grand Ballroom of The Westin Bayshore. Consul General Kwangsuk Lee delivered welcoming remarks to Korean and Canadian guests gathered ahead of Korea’s National Foundation Day on October 3. The program combined a celebration of Korean culture and the Korean community’s contributions with presentations exploring opportunities for bilateral economic and industrial cooperation.

During the reception, Paul Choi, British Columbia’s Parliamentary Secretary for Trade, presented the province’s official proclamation recognizing October as Korean Heritage Month. In his remarks, Choi highlighted the sacrifices and contributions of first-generationHe described it as a symbol of friendship between Korea and Canada, particularly BC. His remarks also connected the legacy of earlier generations with the opportunities ahead for their descendants.

The Korea Trade-Investment Promotion Agency (KOTRA), Soon-Jae Lee, Deputy Director of the Vancouver office outlined Korea’s export profile, manufacturing strengths and prospects for cooperation with Canada. Its presentation covered semiconductors, shipbuilding, artificial intelligence, biotechnology and healthcare, alongside opportunities in energy and logistics. A central theme was how Korea’s technological and manufacturing capabilities could complement Canada’s resources and industrial base.

Samsung Heavy Industries presented its shipbuilding and offshore engineering capabilities. Its presentation featured a next-generation floating liquefied natural gas (FLNG) facility concept and explored opportunities to work with Canada’s LNG industry. The session broadened the view of Korean industry familiar to many through mobile phones and semiconductors, highlighting its role in infrastructure for energy production and transport.

In a toast, the president of the Canada Korea Business Association emphasized the people connecting the two countries’ futures. Families, students, entrepreneurs, veterans and the Korean Canadian community were recognized as foundations of the bilateral relationship. The remarks underscored the longstanding personal ties and trust that support cooperation in trade and security.

The cultural program included “The Korean Way,” a video introducing Korea’s traditions and innovation, and a Hangul calligraphy experience. Guests could have a chosen word or personal wish written in the Korean alphabet to take home as a keepsake, linking Korean culture to their own lives and aspirations. A taekwondo demonstration was also announced as a later part of the program.

The reception brought together the trust built by the Korean community and the industrial capabilities Korea can contribute to the partnership. The pairing of a heritage proclamation with corporate technology presentations illustrated two foundations of the bilateral relationship. The task ahead is to translate the affinity fostered through people and culture into concrete business partnerships and technological exchange.

Reporting from Vancouver

Read the Korean version

Series Overview

The clinical value of artificial intelligence in stroke imaging must be measured by both diagnostic performance and its effect on treatment times and patient recovery. Rapid identification of an intracranial hemorrhage or large-vessel occlusion produces limited benefit if clinicians see the alert late or if interhospital transfer and mechanical thrombectomy are delayed.

Patients with suspected acute ischemic stroke generally receive non-contrast computed tomography to rule out hemorrhage, followed by CT angiography to identify an occluded artery. CT perfusion imaging or magnetic resonance imaging may then be used to distinguish irreversibly damaged brain tissue from tissue that could still be saved. Once a large-vessel occlusion is confirmed, the hospital must activate a thrombectomy team or transfer the patient to a center capable of performing the procedure.

AI can prioritize emergency scans, identify suspected lesions, calculate their location and volume, and send alerts to clinicians. These products should be evaluated across four separate areas: image analysis, regulatory clearance, clinical workflow integration and patient outcomes. Regulatory approvals and hospital installations show market access, but they do not establish faster treatment or better functional recovery at 90 days.

A research team led by Jeffrey Saver, a professor at the University of California, Los Angeles, analyzed 1,287 patients enrolled in five randomized trials. The study found that the benefit of thrombectomy declined as the time between symptom onset and treatment increased.

When treatment began three hours after symptom onset, the absolute difference in the likelihood of achieving a lower disability score was 39.2%. That figure fell to 30.2% at six hours and 15.7% at eight hours. Stroke AI must therefore be assessed by how much it accelerates reperfusion treatment, as well as by how quickly it processes an image.

JLK and Global Rivals Require the Same Evaluation Standards

South Korea’s JLK has assembled 12 stroke imaging applications under its MEDIHUB STROKE portfolio. Its products analyze CT, CT angiography, CT perfusion and MRI scans for conditions including intracranial hemorrhage, large-vessel occlusion and acute cerebral infarction.

According to JLK’s business report filed in March 2026, JLK-LVO, JLK-CTP, JLK-PWI, JLK-ICH and JLK-SDH have received 510(k) clearance from the US Food and Drug Administration. Seven products, including JLK-DWI, have passed Japan’s medical-device review process. In South Korea, JLK-DWI, which analyzes ischemic lesions on diffusion-weighted MRI, has entered the non-reimbursed clinical pathway for innovative medical technologies.

These clearances permit the company to market products in the United States and Japan. FDA 510(k) clearance establishes substantial equivalence to a legally marketed device; it does not determine that the new product is clinically superior to competing systems. Multiple product approvals also do not demonstrate that the applications operate together as one integrated hospital workflow.

JLK’s image-analysis performance has been examined in South Korean studies. In a multicenter study involving 595 patients, JLK-LVO achieved sensitivity of 85.8% and specificity of 96.9% in detecting large-vessel occlusion. Less-experienced physicians also improved their detection sensitivity when using the AI system.

A 2026 South Korean real-world comparison analyzed images from 176 patients with suspected stroke using JLK-LVO and RAPID CTA from US-based RapidAI. Both systems achieved an area under the receiver operating characteristic curve of 0.93. At a matched sensitivity of 83%, specificity was 96% for JLK-LVO and 89% for RAPID CTA.

The results do not establish that JLK-LVO is generally superior to RAPID CTA. The study was a retrospective analysis conducted at a single stroke center, and a JLK researcher was among the co-authors. Both systems sometimes classified severe arterial stenosis as an occlusion and missed distal occlusions when collateral circulation was well developed.

Another study compared JLK-DWI and RAPID MRI with manual lesion measurements in diffusion-weighted MRI scans from 414 patients. RAPID MRI failed to detect lesions in 61.4% of cases, compared with 1.9% for JLK-DWI. Infarct volumes calculated by both systems correlated strongly with manual measurements.

JLK-DWI performed better in detecting small lesions in this study. The research, however, compared specific versions of the two products and included participation by JLK personnel. It did not determine whether detecting more lesions changed decisions on thrombolysis or thrombectomy or improved functional recovery at 90 days.

A separate CT perfusion study illustrates why hospitals should not use AI-generated measurements as the sole basis for treatment decisions. Researchers compared JLK-CTP and RAPID in 327 South Korean patients and found differences in the volumes calculated for the ischemic core and hypoperfused tissue.

Both systems tended to overestimate the volume of irreversibly damaged tissue and showed weaker agreement with early follow-up MRI in patients with small infarcts. An overestimated ischemic core could exclude a potentially eligible patient from thrombectomy. Clinicians must therefore examine the original images, neurological findings and time of symptom onset alongside the AI-generated measurements.

JLK-NCCT, which received US clearance in 2026, analyzes non-contrast CT scans for suspected intracranial hemorrhage and large-vessel occlusion and sends mobile alerts. FDA documentation defines it as a prioritization and notification tool rather than a diagnostic device. A negative result cannot rule out disease, and standard care may still require CT angiography or MR angiography.

JLK has established a broad imaging portfolio, regulatory clearances in the United States and Japan, and several direct product comparisons. Its published research, however, remains concentrated on lesion detection and volume measurement. Publicly available multicenter evidence remains limited on overseas use, clinician adoption, alert-response times, transfer times, repeat imaging, thrombectomy initiation and functional recovery at 90 days.

US-based Viz.ai detects suspected large-vessel occlusions, sends mobile alerts and allows referring hospitals and thrombectomy centers to share images and patient information. RapidAI, also based in the United States, combines occlusion detection and perfusion analysis with transfer decisions. UK-based Brainomix supplies non-contrast CT, CT angiography and CT perfusion products to healthcare providers across Europe.

JLK offers a broad range of CT and MRI applications. Viz.ai and RapidAI have accumulated more evidence on mobile collaboration, care coordination and interhospital transfers, while Brainomix has expanded its experience within European health systems and national programs. JLK has developed imaging technology and regulatory access that allow comparison with global competitors, but it has released less evidence on overseas hospital operations and patient outcomes.

Faster Treatment Must Lead to Better Recovery at 90 Days

A 2026 study by researchers at the US Methodist Health System reviewed 12 studies of AI-based stroke triage and notification systems. Published in Cardiovascular and Interventional Radiology, the journal of the Cardiovascular and Interventional Radiological Society of Europe, it found that AI implementation was associated with a 17.12-minute reduction in the time from hospital arrival to groin puncture for thrombectomy.

The time to the first thrombectomy pass fell by 26.55 minutes, while the time to revascularization declined by 14.55 minutes. For patients transferred from another hospital, the time between arrival and departure from the referring hospital was shortened by 36.8 minutes.

Only one of the 12 studies was a clinical trial; the remaining 11 were observational studies. The researchers concluded that randomized controlled trials were still needed. A separate published review also found shorter times to groin puncture and reperfusion after AI implementation but no statistically significant improvement in functional independence at 90 days or mortality.

Most real-world studies compare performance before and after an AI system is introduced. Hospitals may revise emergency procedures, imaging protocols, transfer rules and staff training during the same period. This makes it difficult to separate the effect of the algorithm from broader operational improvements.

A prospective multicenter study in Switzerland found that AI implementation reduced the time from imaging to a transfer request by about seven minutes and shortened the interval from emergency department arrival to groin puncture by approximately 31 minutes. The proportion of patients receiving CT angiography at referring hospitals rose from 28% to 81%, while repeat imaging at receiving centers fell from 79% to 42%.

The study’s primary endpoint—the time from patient arrival to the transfer request—did not improve significantly. The findings therefore do not show that AI alone accelerated the transfer decision. Wider use of CT angiography, fewer repeat scans and changes in interhospital workflow also contributed to faster treatment.

Tessa Cook, a professor of radiology at the University of Pennsylvania and chair of the committee that developed the first imaging AI practice parameter from the American College of Radiology and the Society for Imaging Informatics in Medicine, said the guidance covers the entire process from selecting an AI product to performance monitoring and continuous quality improvement. She has emphasized that hospitals must continuously manage the deployment, use and updating of rapidly changing clinical AI.

The same standards should apply to JLK and its overseas competitors. Performance can vary across hospitals as patient populations, scanners and imaging protocols change. Healthcare providers should test products on local data before deployment and continuously monitor false positives, false negatives, processing times and performance changes following software updates.

In South Korea, the effectiveness of stroke AI will depend on coordination between regional hospitals and the country’s cardiovascular and cerebrovascular disease centers. An alert from a regional hospital will not accelerate treatment if an ambulance, receiving bed or thrombectomy team is unavailable. Hospitals should report alert-response times, transfer-acceptance times, repeat imaging rates and thrombectomy rates, rather than relying only on installation numbers and scan volumes.

Patient recovery can be measured using the modified Rankin Scale, or mRS, 90 days after a stroke. The scale grades disability from 0 to 6, with 0 indicating no symptoms and 6 representing death. Scores of 0 to 2 generally indicate functional independence in daily life.

JLK has built an imaging portfolio and diagnostic performance that can be compared with leading international companies. Its publicly available evidence, however, remains focused on regulatory clearances in the United States and Japan and diagnostic studies conducted primarily in South Korea. Independent multicenter research on overseas clinical use, interhospital transfers, treatment times and functional recovery at 90 days remains insufficient to establish its clinical competitiveness in global markets.

Viz.ai, RapidAI and Brainomix have published more real-world evidence on lesion detection, clinician alerts and treatment-time reductions than JLK. However, evidence that their faster workflows consistently improve functional recovery, mortality and healthcare costs across multiple independent institutions also remains limited.

Domestic and international companies should therefore be evaluated under the same measures: the number of active clinical sites, interhospital transfer times, treatment speed, functional recovery at 90 days, mortality and changes in medical costs. These measures distinguish product deployment from demonstrated clinical value.

The role of stroke AI does not end when it marks an occluded artery. Its clinical value is established when alerts accelerate specialist review, interhospital transfer and thrombectomy and reduce long-term disability. The international competitiveness of South Korean medical imaging AI companies will depend on their ability to demonstrate this entire process through independent real-world evidence.

Part 6 will examine how the DICOM standard for medical-image descriptions and metadata enables hospitals to operate multiple AI applications safely.

Read the Korean version

NC, jointly led by Kim Taek-jin and Park Byung-moo, said Friday that the CBT for Astrae Oratio has begun and will run through Oct. 6. The game is being developed by Dynamis One.

The five-day test is available to selected participants on mobile devices. Players will be able to experience key elements of the game, including its main storyline, battles and interactions with characters.

Set in Tokyo, Astrae Oratio combines modern urban settings with a fantasy narrative centered on magic. Three main storylines are available during the CBT: “Tokyo, Civil Servants, and Mages,” “The Sparkling Days of Minato Ward Mages,” and “The Night of the Missing Mage and Urban Legend Hunters in Shinjuku Ward.”

The scenarios unfold across different areas of Tokyo, allowing players to gradually explore the game’s world and meet its characters. Two additional mini-stories presented in trailer-style formats are also included in the test.

Combat is another major part of the CBT. Battle quests set in Minato and Shinjuku wards are offered across four difficulty levels — Normal, Hard, Very Hard and Challenge.

Players can also take on “Score Trial,” a boss-focused mode that introduces various battle conditions, as well as “Simulation Room,” a dungeon designed to provide resources for character progression.

Beyond combat, the game features a character interaction system called “Free Time.” Players can spend time with characters, learn more about their daily lives and increase their relationship levels. The CBT also allows players to experiment with character costumes and voice content.

NC plans to use gameplay data and feedback collected during the test to identify areas for improvement and incorporate the findings into the game’s ongoing development and preparations for its global launch.

More information about the Astrae Oratio CBT is available through the game’s official website and social media channels.

Series Overview

Part 1 | Why Does AI Get Things Wrong So Convincingly?
Part 2 | Why the OXMV Loop and Evidence Ledger Are Necessary
Part 3 | Mapping Knowledge for AI with Ontology
Part 4 | OpenAI to End ChatGPT’s Custom GPT Era on December 11, 2026
Part 5 | Taking Aim at ChatGPT’s Chronic “Hallucination” Problem

By Manripungchwi Jung Yeon-tae

Generative artificial intelligence may be entering a new phase of competition. Until now, much of the race in AI has focused on which systems can generate more fluent, capable and sophisticated answers. But as AI moves deeper into journalism, law, medicine, finance, research and corporate decision-making, another question is becoming increasingly important: Can AI verify what it is saying—and recognize when the evidence does not justify a confident answer?

An OXMV-based GPT, built around a verification-centric framework called OXMV, is attempting to address that problem from a different direction. Its most striking principle is surprisingly simple: “Do not automatically trust even the premise of the user’s question.”

An AI That Questions the Question

Consider a user asking, “Why did A cause B?” The question already contains an assumption: A caused B. Or consider, “A study proved that this treatment works. Why is it effective?” That question may contain several assumptions—that the study exists, that it actually reached the claimed result, and that its findings justify the word “proved.”

OXMV is designed not to automatically promote such assumptions into verified facts. Instead, potentially material factual assumptions contained in a prompt can be separated into what the system calls Prompt Premises and evaluated before they are allowed to serve as factual foundations for subsequent reasoning. The governing principle is: Prompt Presence ≠ Factual Authority.

In other words, simply because a claim appears in a user’s question does not make it true. This distinction may sound obvious, but it addresses one of the more subtle failure modes of conversational AI: a model can accept a false premise embedded in a perfectly natural question and then produce an eloquent explanation of something that never happened.

Breaking Answers Down Into Verifiable Claims

OXMV also treats sentences not merely as blocks of text, but as collections of potentially independent claims. A single sentence may contain a factual assertion, a number, a comparison, a causal relationship, an interpretation and a recommendation. Under the OXMV architecture, these can be decomposed into Atomic Claims so that each material proposition can be evaluated separately.

Important boundaries—such as time, population, scope, conditions, exceptions, quantifiers and modality—are intended to remain attached to the claim during verification. That matters because hallucination does not always take the form of an entirely fabricated fact. Sometimes the error is subtler.

“An association was observed” can become “A causes B.” “May improve” can become “has been proven to improve.” “Some patients” can become “patients.” “A result observed under specific conditions” can become a universal conclusion.

OXMV treats such semantic strengthening as something that must be controlled rather than dismissed as stylistic rewriting.

A Source Is Not the Same as Evidence

Another central OXMV principle is: Source ≠ Evidence. Finding a paper, government report, court decision or official webpage does not automatically establish that the source supports the claim being made. OXMV therefore separates several questions that are often collapsed into one.

Does the source actually exist? Is it the correct source and version? Was the relevant original material actually accessed? What does the relevant passage say in context?

Does that passage directly support the claim? Does it support the same population, time period, scope and conditions? Is the evidence sufficient for the strength of the claim? Is the source current?

Are supposedly independent sources actually derived from the same underlying material? In this architecture, retrieval is not verification. A search result, snippet, abstract, press release or AI-generated summary is not automatically equivalent to examining the underlying evidence.

A Correct Calculation Does Not Make the Input True

OXMV applies the same distinction to numbers. Suppose a user provides an incorrect revenue figure, but the AI performs a mathematically flawless calculation using that number. Was the calculation wrong? Not necessarily.

Was the resulting real-world conclusion verified? Also not necessarily. OXMV explicitly separates these issues through the principle: Calculation Correctness ≠ Input Truth.

The framework similarly distinguishes percentages from percentage points, absolute from relative change, counts from rates, and actual results from estimates, forecasts, scenarios and targets. It also seeks to prevent statistical association from being automatically transformed into causation.

Not Every Error Is an “AI Hallucination”

One of OXMV’s more distinctive features is its attempt to separate different origins of error rather than labeling every incorrect statement a hallucination. Its architecture distinguishes categories including:

PROMPT_PREMISE_ERROR — an error already present in the user’s factual premise.
SOURCE_ERROR — an error already present in a retrieved source.
ANALYTICAL_ERROR — an error arising in calculation, statistics, logic, causal analysis or domain application.
HALLUCINATION — unsupported factual or semantic content newly introduced during generation.
MIXED_ERROR — multiple error origins affecting the same conclusion.
UNKNOWN_OR_UNRESOLVED — cases where the available evidence does not justify determining the origin.

This distinction has practical consequences. If a retrieved source itself contains an incorrect number and the AI accurately reports, “This source states X,” the upstream error and the model’s generation behavior are not necessarily the same problem.

But if the model takes that source and adds an unsupported causal conclusion, removes an important condition, expands a finding to a broader population, or converts uncertainty into certainty, that additional transformation may constitute a separate generation error.

The OXMV Loop: Verification Does Not End When the Answer Is Written

The heart of the architecture is what can be described as a closed-loop verification process. Many systems can retrieve information before generating an answer. OXMV attempts to go further.

If new evidence or counterevidence appears, a source changes, a calculation is corrected, or an important assumption is revised, the system is designed to identify which claims depend on the changed information. It can then determine which claims require reverification, reassess the permitted strength and scope of those claims, regenerate the relevant output, and inspect the generated answer again.

In simplified form:

Verify → Generate → Inspect the Generated Output → Detect Change or Error → Trace Dependencies → Reverify → Regenerate

The important idea is that verification is not intended to be a one-time checkpoint before generation. The generated answer itself becomes another object that must remain within the boundaries authorized by the evidence.

The Evidence Ledger: Keeping Track of What the AI Actually Knows

If the OXMV Loop represents the process, the Unified Verification Ledger represents the state-control architecture behind it. The framework is designed to maintain relationships among claims, evidence, sources, counterevidence, gaps, unresolved verification work, mutations, dependencies, output constraints and verification snapshots.

One especially important principle is that “Verified” is not treated as a permanent label. A law can change. A scientific paper can be corrected or retracted. A clinical guideline can be updated.

A market price can change within seconds. Software and AI models can move to new versions. A claim verified under one set of conditions therefore may not remain verified indefinitely under another. OXMV treats verification as context-bound to factors such as time, source version, scope, population and conditions.

How Is This Different From ChatGPT, Gemini or Claude?

The distinction requires care. General-purpose AI systems such as ChatGPT, Gemini and Claude are themselves evolving rapidly. Their developers continue to improve reasoning, retrieval, tool use, grounding, factuality and other mechanisms intended to reduce errors. It would therefore be inaccurate to claim that mainstream AI systems “do not verify information.”

The more meaningful distinction is architectural emphasis. General-purpose frontier AI systems are designed to solve an extremely broad range of tasks. OXMV, by contrast, places an explicit verification and state-control architecture around the question: What is the maximum claim that the currently available evidence actually authorizes the AI to make?

The difference can be summarized this way:

“If conventional generative AI primarily asks, ‘What answer should I generate?’, OXMV adds another question: ‘After checking the evidence, how much of that answer am I actually justified in saying?’”

This does not by itself establish that OXMV is more accurate than ChatGPT, Gemini or Claude. That is an empirical question requiring controlled comparative testing. But it does represent a different design philosophy: verification is treated not merely as a supporting feature, but as a control layer over generation.

Hundreds to Thousands of Hallucination-Inducing Test Patterns

According to the developer, OXMV has been tested repeatedly against hundreds to thousands of hallucination-inducing question patterns, including comparative tests involving general-purpose AI systems. The adversarial cases are designed to include such traps as false numerical premises, nonexistent sources, incorrect descriptions of real sources, outdated premises presented as current facts, causal assumptions embedded in questions, unsupported comparisons, and forecasts presented as actual results.

These are not merely hypothetical categories. OXMV’s own adversarial framework explicitly defines test families for false numeric premises, nonexistent-source premises, wrong-source descriptions, stale premises, causal presuppositions, comparative presuppositions and forecast-as-actual errors. The developer says that, to date, the OXMV-based GPT has passed the prepared hallucination-inducing internal tests.

That claim, however, should be understood within its proper evidentiary boundary. It is a report of developer-run testing, not yet equivalent to independent third-party blind validation. Importantly, OXMV itself does not claim that hallucinations have been eliminated with 100 percent certainty.

Its own framework explicitly distinguishes development and regression testing from independent blind or holdout evaluation and states that its “Target 99” concept is a development objective—not a measured guarantee of 99 percent performance. An important next step would therefore be independent testing in which OXMV and systems such as ChatGPT, Gemini and Claude are evaluated under identical conditions, with model versions, prompts, tools, scoring rules and test sets controlled in advance.

Why Verification-Centric AI Matters Now

The significance of such an approach goes beyond whether an AI gets an individual trivia question right or wrong. AI systems are increasingly being considered for tasks involving journalism, law, medicine, finance, taxation, scientific research, corporate strategy and public policy. The cost of error in these fields can be dramatically different from the cost of an incorrect restaurant recommendation or casual factual mistake.

As AI enters higher-stakes environments, users may increasingly need answers to questions beyond “What did the AI say?” They may need to know: What assumptions did it rely on? Which sources were actually checked? Did those sources really support the claim?

Was counterevidence considered? Which facts remain unverified? What conditions or exceptions apply? If a source changes tomorrow, which conclusions must be rechecked?

This is where verification architecture could become economically important.

Potential Impact on Journalism

Journalism provides an especially clear example. Suppose a reporter asks an AI, “Why did Company A conceal the incident?” The question already presupposes both that an incident occurred and that Company A concealed it.

If those assumptions are not established, an AI that accepts the question literally could produce an articulate explanation of motives for an event or concealment that has not been verified. A verification-centric system would first identify the embedded premises.

For newsrooms, this could make Claim–Evidence–Source traceability as important as the quality of generated prose. The issue becomes even more important when AI-generated errors enter articles, social media and search results and are later recycled as apparent “sources” for subsequent AI systems.

Law, Medicine and Finance Could Be Even More Significant

The potential consequences become greater in high-risk domains. In law, the existence of a statute is different from whether it applies to a particular case. In medicine, evidence that a treatment works in a study population is different from determining that it is appropriate for a specific patient.

In finance, a forecast is not an actual result. In taxation, the existence of a deduction is not the same as a particular taxpayer being eligible for it. In safety engineering, “no reported incident” does not mean “no risk.”

OXMV V3.4 contains separate verification paths for legal and regulatory analysis, scientific and medical evidence, finance and investment, accounting and tax, policy, and safety.

The Next AI Race May Not Be About Knowing More

As AI systems become more capable, one of their most valuable abilities may paradoxically be knowing when not to make a stronger claim. Not inventing a missing fact. Not treating “not found” as “does not exist.” Not assuming that the absence of counterevidence proves a proposition true.

Not mistaking a correct calculation for proof that its inputs were correct. Not converting a hypothesis into reality. Not converting association into causation. And when something cannot be verified, preserving that uncertainty instead of silently filling the gap.

That leads to perhaps the simplest description of what OXMV is trying to build:

“Not an AI that simply generates more answers, but an AI designed to say only as much as the evidence permits.”

Whether this architecture can consistently outperform mainstream general-purpose AI systems remains a question for controlled, independent evaluation. But the larger question raised by OXMV may be even more important: Will the next stage of generative AI competition be defined not only by intelligence, speed and fluency—but by verifiable trust?

About the Author

Manripungchwi Jung Yeon-tae is a professional-knowledge AI planner, designer, and GPT development specialist. He currently serves as President of the Construction Industry Information Institute and develops professional GPTs across construction, law, taxation, finance, management, medicine, science, history, and philosophy. His research focuses on verified professional-knowledge AI combining ontology, XAI, the OXMV Loop, and the Evidence Ledger.

Korean version available here

The brokerage has moved beyond conventional dollar-denominated bonds by issuing foreign-currency hybrid capital securities for the first time among South Korean securities firms. The move adds a capital-eligible instrument to its funding portfolio as the company seeks to diversify both funding sources and maturities.

On Sept. 30, Hana Securities issued $300 million in U.S. dollar-denominated hybrid capital securities. The notes have a 30-year maturity and are callable after five years under a 30NC5 structure. S&P assigned a BBB rating to the securities.

The offering attracted orders of $1.7 billion from more than 120 institutional investors worldwide, equivalent to nearly six times the issue size. The coupon was set at 6.50%, or 169 basis points above the five-year U.S. Treasury yield, 30 basis points tighter than the initial price guidance.

Foreign-Currency Funding Expands Across Maturities

The latest transaction extends a foreign-currency funding strategy that Hana Securities has been building over the past several years.

Since entering the overseas public bond market in 2024, the brokerage has continued to tap global investors. Its funding structure has expanded from three-year dollar bonds to five-year senior notes and now to 30-year hybrid capital securities.

Hana Securities’ foreign-currency bonds stood at about 1.97 trillion won as of the end of June, up from 623.4 billion won at the end of 2024. The company has also increased its global medium-term note program limit from $2 billion to $5 billion, giving it greater flexibility to issue bonds in international markets.

Earlier this year, Hana Securities issued $300 million in five-year senior notes. The deal attracted more than $3.3 billion in orders from 168 institutional investors, demonstrating continued demand from overseas investors.

The latest hybrid capital securities differ from conventional senior bonds in that they can be recognized as regulatory capital under applicable requirements. Hana Securities expects the transaction to strengthen its capital adequacy while providing additional capacity for business expansion.

New Funding Capacity Supports Corporate Finance

The expansion of foreign-currency funding comes as Hana Securities increases its focus on corporate finance and productive investment.

The brokerage has been diversifying its funding channels while directing additional financial capacity toward businesses including corporate financing, equity capital markets, acquisition financing and venture investment.

Its short-term note business has also emerged as another funding channel. Hana Securities began the business this year and has been using the funds raised to support areas including corporate lending and venture investment.

The company has also reorganized parts of its investment banking operations to strengthen its focus on productive finance, with greater emphasis on providing capital to companies and growth industries.

Global IB and Venture Capital Among Growth Areas

Hana Securities plans to use the additional financial capacity created by the hybrid securities issuance to expand global IB, corporate finance, venture capital investment and other growth businesses.

The company is also seeking to build longer-term relationships with global institutional investors by regularly accessing overseas capital markets and broadening the range of foreign-currency instruments it can issue.

The latest transaction adds a long-term capital instrument to a funding structure that already includes senior foreign-currency bonds and other financing channels.

Kim Jung-hoon, head of the FICC division at Hana Securities, said the company secured stable investor demand through continued communication with major global investors ahead of the issuance.

“Following the first foreign-currency hybrid capital securities issuance by a Korean securities firm, we will continue to strengthen our capital-raising capabilities by building long-term relationships with global investors,” Kim said.

Samsung said October 1 that a gala dinner was held at the British Museum in London on September 30 to mark the opening of “Korea: 2,000 Years of Creativity,” a special exhibition jointly organized by the National Museum of Korea, the National Museum of Modern and Contemporary Art, Korea, and the British Museum.

The exhibition marks the British Museum’s first major exhibition of Korean art in 42 years. It also represents the third stop of the overseas tour of the KH Collection, following exhibitions in Washington, D.C., and Chicago.

At the heart of the exhibition are works donated to the Korean government by the family of Lee Kun-hee in 2021. The donation included approximately 23,000 cultural artifacts and artworks, including national treasures and treasures, collected by Lee over decades.

The donated works are now housed at institutions including the National Museum of Korea and the National Museum of Modern and Contemporary Art and have been displayed at exhibitions across Korea. The overseas tour is giving international audiences an opportunity to view some of the collection’s major works.

The Lee Kun-hee Collection spans a broad range of periods and genres, from ancient cultural artifacts and Buddhist art to Goryeo celadon, Joseon-era paintings and crafts, and modern and contemporary Korean art.

Among the works being presented in London are “Inwang Jesaekdo” by Jeong Seon, “Chuseongbu-do” by Kim Hong-do and “Paradise” by Paik Nam-soon. The exhibition also features 10 designated national treasures and 13 treasures.

The exhibition traces the development of Korean creativity across approximately 2,000 years rather than focusing on a single artistic period. Through works ranging from ancient artifacts and Goryeo ceramics to Joseon paintings and modern art, visitors can explore the evolution of Korean artistic traditions.

The London exhibition follows the collection’s first overseas showing at the Smithsonian’s National Museum of Asian Art in Washington, D.C., in November last year. About 80,000 people visited the exhibition. The Chicago exhibition, held from March through July this year, attracted approximately 115,000 visitors, bringing the combined attendance in the two U.S. cities to around 195,000.

Samsung Electronics Executive Chairman Lee Jae-yong highlighted the importance of preserving and sharing cultural heritage at the gala dinner.

“My family has upheld the principle across generations that art and culture should be preserved and shared,” Lee said. “I hope this exhibition will provide an opportunity to understand 2,000 years of Korean creativity and cultural heritage.”

The Lee Kun-hee Collection has drawn attention not only for the scale of the original private collection but also for its transformation into a public cultural asset. Following the 2021 donation, works that had once been privately held became accessible to the public through national museums and exhibitions.

The overseas tour has further expanded that role by bringing Korean cultural heritage to audiences outside the country. The London exhibition comes as international interest in Korea has expanded from contemporary popular culture such as K-pop and Korean cinema to the country’s traditional arts and cultural history.

By presenting works from ancient Korea through the modern era in one exhibition, the British Museum show offers international visitors a broader view of the historical foundations and artistic development of Korean culture.

The London exhibition represents another step in the international journey of the Lee Kun-hee Collection, transforming a personal passion for collecting into a cultural legacy that is being shared with audiences in Korea and abroad.

The company said on October 1 that it will build a chiller manufacturing plant in Windsor, Virginia, with production scheduled to begin in the first half of next year. The facility will cover approximately 32,000 square meters of floor space on a 170,000-square-meter site and will primarily produce air-cooled chillers for AI data centers.

The investment is aimed at establishing a local supply base closer to North American data center customers. Virginia, known as a major U.S. data center hub, has attracted large-scale facilities operated by major technology companies, making the region an important market for cooling equipment.

Demand for cooling systems is rising alongside the expansion of AI data centers. High-performance GPUs and CPUs generate significantly more heat than conventional data center equipment, making efficient and stable thermal management increasingly important. According to Omdia, global data center power capacity is expected to rise from 226 gigawatts this year to 420GW by 2030, with North America accounting for about 60% of the total.

LG Electronics is also expanding its domestic production infrastructure. The company plans to add air-cooled chiller production lines and upgrade manufacturing equipment at its plants in Pyeongtaek and Changwon.

Total investment in the new U.S. facility and the expansion and modernization of production lines in Korea is expected to reach approximately 150 billion won ($110 million).

Industry observers say cooling infrastructure is becoming a critical component of AI data center expansion as operators seek to manage the higher heat loads generated by advanced computing systems. Establishing local production in North America is expected to help LG Electronics improve supply stability and respond more quickly to growing demand in the region.

LG Electronics is also broadening its data center cooling portfolio beyond chillers. The company plans to expand its business around liquid-cooling solutions, including coolant distribution units (CDUs) and cold plates that directly absorb heat generated by semiconductor chips.

The company is pursuing a “Chip to Chiller” strategy that connects cooling technologies from the chip level to the chiller, allowing it to offer integrated thermal management solutions for data center operators.

“We will continue to strengthen our supply capabilities for cooling solutions, a core infrastructure component of AI data centers, and actively respond to rapidly growing market demand,” said Lee Jae-sung, president of LG Electronics’ Eco Solution (ES) Company.

San Francisco-based AI company Use.ai said on October 1, 2026, that it had surpassed one million monthly active users less than a year after launch. Its workspace provides access to more than 10 AI models, including GPT, Claude, Gemini, Grok and DeepSeek. Users can switch models within a conversation while retaining its context, according to the company.

Use.ai offers a free Answer Engine based on open-weight models that it configures and optimises, with more advanced models available through a paid subscription. Users can draw on different models as a research, writing or planning task develops. The platform is designed to reduce the need to move between separate services and re-enter conversations or preferences.

Co-founder Ihor Herasymov previously worked in mergers and acquisitions and technology investment. According to the company, he went on to launch seven businesses and grew a technology company to more than 100 employees and $3 million in annual revenue within two years. His experience using multiple AI platforms drew his attention to how difficult it was to carry conversation histories and personal preferences from one service to another. “I don’t want to lose that context whenever a better model appears,” Herasymov said, describing the idea behind Use.ai as one account that keeps context available across models.

In a company survey of more than 5,000 users, 42% said they switched models for better quality or deeper reasoning, while 36.5% said their choice depended on the task. A separate survey of more than 30,000 users found that 54% were either using AI for the first time or had tried it only a few times. Use.ai is developing automatic routing to select a model suited to each task, reducing the need for users to compare models themselves. The company did not disclose how respondents were recruited or how representative the surveys were of its wider user base.

Use.ai’s next goal is to extend the platform from answering questions to carrying out tasks. With a user’s permission, a proposed personal assistant could connect to calendars and other apps to help arrange appointments or prepare purchases based on the user’s schedule and preferences. For example, the company envisages an assistant checking salon availability, making a booking and setting a reminder. It also describes a shopping assistant that could compare running shoes and prices, then seek the user’s confirmation before making a purchase. These are examples of planned capabilities, rather than features the company says are available today.

“Our users shouldn’t need to follow model releases, compare benchmarks or decide which model is best for every task,” Herasymov said. Use.ai plans to build on model switching and shared context by coordinating AI models and connected apps to complete more of a user’s request.

New York-based finance technology company Maximor said on October 1, 2026, that it had changed its name to Hyphenate as it expanded its platform across the office of the chief financial officer. The company reported that revenue grew 86-fold over the past 12 months. It did not disclose the revenue figure used as the basis for that comparison or the size of its customer contracts.

Hyphenate said its platform now covers five areas: order-to-cash, treasury, general ledger accounting and close, procure-to-pay, and reporting and insights. Customers use an average of six modules, and 40% expand their use within the first year of a contract, according to the company. The platform is designed to connect with existing enterprise resource planning, banking and payroll systems so finance teams can work across those systems without replacing them.

Ramnandan Krishnamurthy, Hyphenate’s co-founder and chief executive, said customers often begin with one finance workflow before extending the platform to others. The company says its AI agents record the work they perform and refer decisions requiring human judgment to finance staff. The extent of automation and the approval controls will depend on each company’s processes and policies.

Yameen Sarwar, finance transformation lead at customer Mini Melts USA, said the company uses Hyphenate’s agents in its SAP environment for tasks including orders, invoicing and refunds. Mini Melts is expanding its use of the platform into treasury and lease management, Sarwar said. Hyphenate’s longer-term aim is to keep financial information current throughout the month, reducing the time between business activity, reconciliation and management reporting.

The statement comes amid growing market attention over reports that Solidigm could seek outside capital or pursue a separate listing in the United States as demand for high-capacity enterprise SSDs (eSSDs) rises alongside investment in artificial intelligence data centers.

“We are reviewing various options to strengthen Solidigm’s competitiveness, but our position that nothing has been finalized remains unchanged,” SK hynix said.

The company added that any funding decision would take into account market conditions, investment timing, its financial position and the potential economic impact on existing SK hynix shareholders.

Reports Point to Possible U.S. IPO

Reuters reported on Sept. 25, citing people familiar with the matter, that Solidigm was considering a U.S. IPO as early as 2027 and had held discussions with investment banks over potential underwriting roles.

The report cited a possible valuation of up to $150 billion and potential proceeds of around $15 billion. The size and timing of any transaction remain at an early stage and could change depending on market conditions.

Separate reports have also raised the possibility that Solidigm could seek around 5 trillion won in pre-IPO investment.

SK hynix previously responded that it was considering multiple options to enhance Solidigm’s competitiveness but that no specific plan had been decided. The company has maintained the same position in disclosures filed with the U.S. Securities and Exchange Commission.

The reports have also raised concerns in South Korea over a potential separate listing of Solidigm and its implications for SK hynix shareholders.

AI Data Centers Drive Demand for High-Capacity eSSDs

Solidigm was established following SK hynix’s acquisition of Intel’s NAND and SSD business. SK hynix agreed in 2020 to acquire the business for approximately $9 billion and subsequently carried out the transaction in stages.

The U.S.-based subsidiary focuses primarily on enterprise SSDs and has expanded its presence in high-capacity storage products for data centers.

Demand for enterprise storage has increased as global technology companies expand AI infrastructure. The rapid growth of data generated and processed by AI workloads is driving demand for high-capacity storage alongside high-performance computing and memory products.

Solidigm has been expanding its portfolio of high-capacity eSSDs based on quad-level cell, or QLC, NAND technology for AI and data center applications.

For SK hynix, the business complements its growing portfolio of AI-related products, which includes high-bandwidth memory, server DRAM and enterprise storage.

The expansion of Solidigm’s business, however, also requires continued investment in production capacity, product development and technology.

SK hynix Weighs Internal Funding and External Capital

SK hynix said it has the financial capacity to fund Solidigm internally but does not regard the use of its own capital as the only option.

The company is simultaneously investing across several businesses, including HBM, server DRAM, NAND flash and enterprise SSDs, making capital allocation an important consideration in determining how additional Solidigm investment would be funded.

Potential options could include internal funding or the use of external capital. However, SK hynix stressed that no particular financing method has been finalized.

The company said the choice of funding source would be considered in light of market conditions, the timing of investment, its financial position and its broader capital expenditure plans.

SK hynix also said that if external capital is considered, it would compare the option with the use of internal funds and assess the economic impact on existing shareholders.

Shareholder Protection to Be Reviewed if Plans Advance

Market concerns have focused on whether a separate Solidigm listing could dilute the economic interests of existing SK hynix shareholders.

SK hynix said that if plans for a separate listing or external investment move to a more concrete stage, its board would review the financial and business implications as well as measures to protect existing shareholders.

The company emphasized that long-term corporate value and shareholder value would remain central to any decision.

“If we move forward with reviewing a specific plan, we will also examine its impact on existing shareholders and any necessary shareholder protection measures, and provide sufficient explanation,” the company said.

For now, SK hynix maintains that no Solidigm IPO or external fundraising plan has been finalized. The company continues to review funding options as it expands its AI memory and enterprise storage businesses.

The company said on Oct. 1 that U.S. sales of bibigo gimbap increased 59% year-on-year during the January-July period.

Since entering Kroger in December 2024, bibigo gimbap has expanded into major retailers including Albertsons, Food Lion, Cub and Meijer. The products are now available through 10 mainstream U.S. retail chains.

The number of stores carrying bibigo gimbap has also increased sharply, rising 145% from 1,924 locations last year to 4,734.

Positioning Gimbap as a Convenient Everyday Meal

CJ CheilJedang sees the U.S. as a key market for expanding gimbap beyond the Korean food category and into everyday meals.

The company believes gimbap fits well with American eating habits, where portable foods such as sandwiches and burritos are widely consumed. Its combination of convenience and familiar ingredients could help lower the barrier for consumers trying Korean food for the first time.

bibigo gimbap also contains 12 grams of plant-based protein, positioning it as an option for consumers looking for a convenient and balanced meal.

CJ CheilJedang is currently selling two varieties in the U.S., Bulgogi and Spicy Bulgogi, and plans to broaden the lineup with additional flavors inspired by Korean ingredients and dishes.

The strategy follows the company’s experience with Korean dumplings, or mandu, which have expanded from a relatively unfamiliar Asian food into a more widely available product in the U.S. market.

Promoting the Identity of Original Korean Gimbap

Alongside its retail expansion, CJ CheilJedang is stepping up consumer marketing to emphasize the Korean origins and identity of gimbap.

The company is running its “OG Recognizes OG” campaign in Los Angeles, using “OG” to represent both “Original” and “Original Gimbap.”

The campaign is designed to introduce American consumers to the authentic taste and cultural identity of Korean gimbap while building awareness of bibigo as the brand expands into mainstream grocery channels.

CJ CheilJedang has also partnered with K-food creator Danny Kim, who has around 9 million followers across TikTok and YouTube, to produce digital content highlighting the convenience and versatility of bibigo gimbap.

“As someone who grew up enjoying gimbap, I was excited to work with bibigo to introduce its taste to more people,” Kim said. “bibigo gimbap is a convenient option when you want to enjoy gimbap but do not have time to make it yourself.”

Expanding Retail Reach and Product Portfolio

CJ CheilJedang plans to continue adding mainstream retail partners while increasing consumer engagement through digital content and collaborations with local creators.

The company will also expand its product portfolio beyond the current Bulgogi and Spicy Bulgogi varieties, introducing new products inspired by representative Korean ingredients and dishes.

“The U.S. has strong growth potential for gimbap, given rising interest in K-food and a food culture that values convenient meals,” a CJ CheilJedang official said. “We will continue expanding mainstream distribution and local consumer marketing to establish bibigo gimbap as a representative K-food that American consumers can enjoy as part of their everyday lives.”

Ascerta’s co-founders. The company, formerly known as Pay-i, has raised $18 million in Series A funding to expand its platform for measuring AI costs, adoption and business value. / Courtesy of Ascerta

Ascerta, formerly known as Pay-i, announced in Bellevue, Washington, on September 30, 2026, that it has raised $18 million in Series A funding led by Dell Technologies Capital. Hitachi Ventures, BGV, Wipro Ventures and existing investors also participated, bringing total funding to $22.9 million. The company will use the capital to expand its platform and go-to-market team as it broadens its focus from tracking AI costs to measuring business value.

The platform connects AI spending and usage with the business outcomes each initiative is intended to deliver. Its aim is to help technology and finance executives decide which projects to expand, adjust or discontinue. That approach addresses a measurement problem: figures such as tokens consumed, agent runs and code generated record activity but do not, on their own, establish a return on investment.

“Companies are counting tokens, lines of generated code, and agent runs, struggling to derive the impact AI has on their business,” said David Tepper, Ascerta’s chief executive and co-founder. He said the company’s tools provide insights specific to individual businesses, employees and use cases. The platform combines those measurements with tools intended to reduce waste and improve returns.

Connecting AI spending with business outcomes

Ascerta was founded in 2024 by former Microsoft employees Tepper, Doron Holan and Erik Winters. It emerged from stealth in May 2025 as Pay-i with $4.9 million in seed funding and an initial focus on AI cost management. The company said its rebranding reflects a broader remit covering AI usage, adoption and financial outcomes.

According to Ascerta, the platform works alongside existing enterprise systems, including internally developed applications, Microsoft’s Copilot suite, Amazon Bedrock AgentCore and Salesforce Agentforce. It also supports coding tools such as GitHub Copilot, Claude Code and Codex. The company says it can connect individual model calls to specific use cases while accounting for charges and adjustments such as hidden fees and enterprise discounts.

Its three products address different parts of that process. Atlas measures adoption, business value and return on investment across workflows and AI portfolios. Forge tracks engineering teams’ use of coding agents to help assess productivity, while Convoy supports organisations managing their own AI capacity and adding workloads.

Customer results require context

Ascerta named Atos, Wipro and global insurance carriers among its customers. Across customers, it reported a 47% improvement in return on AI investment, a 24% reduction in agent launch times and an 86% reduction in wasted AI spending. The announcement did not disclose the sample size, measurement period or calculation methods behind those figures, limiting their usefulness as benchmarks.

“At Atos, our Sovereign Agentic Studios operating model is built on moving agentic AI from pilot to production at global scale,” said Florin Rotar, group chief technology officer and chief AI officer at Atos. He said Ascerta had provided visibility and control to support that expansion. Ascerta said customer use cases include assigning dollar values to AI-powered features and identifying spending associated with failed agent runs, duplicate projects and unauthorised AI use.

Raman Khanna, managing director at Dell Technologies Capital, said enterprises were focusing investment on AI initiatives that deliver measurable business value. Ascerta plans to expand its integrations with enterprise AI tools and develop additional products for improving returns. Its broader proposition will depend on demonstrating how reliably it can link AI activity to financial outcomes, including the baselines and attribution methods used to measure gains.

Solidus Labs announced in New York on October 1, 2026, that digital asset exchange BTSE has selected its HALO Trade Surveillance module to monitor trading across digital assets and tokenized real-world assets. The system is designed to help compliance teams identify potential market abuse as BTSE expands its products and geographic reach. The announcement did not disclose financial terms or a deployment timetable.

According to Solidus Labs, HALO combines trade data with behavioural signals to identify suspicious patterns in real time, including spoofing, wash trading and attempts to artificially push prices higher. BTSE will also use the platform’s Model Testing tools to adjust detection parameters and simulate trading scenarios. The companies said these functions would help tailor alerts to BTSE’s trading patterns and reduce false positives.

“Since we've expanded our services to serve users in many different regions and countries, it's become increasingly difficult to monitor each and every transaction for potential market abuse and manipulation,” said Jeff Mei, chief operating officer at BTSE. He said the technology would help the exchange monitor activity more efficiently. HALO’s Market Inspection function will also give analysts broader market context beyond BTSE’s own trading activity to support investigations.

The agreement covers surveillance for digital asset and tokenized real-world asset markets, according to the announcement. BTSE also licenses exchange technology to third-party operators through its white-label business. The release did not specify whether all exchanges using that infrastructure would be covered by the surveillance programme.

Diana Xu, head of Asia-Pacific at Solidus Labs, said the company would support BTSE’s expansion into new asset classes, including tokenized real-world assets. The announcement centres on strengthening detection and investigation capabilities, but provides no measured results on alert accuracy or reductions in market abuse. Those outcomes will be important in assessing the system’s effectiveness as BTSE broadens its trading operations.

LS Electric said Wednesday that it signed a $133.41 million contract to supply 345kV ultra-high-voltage transformers for a North American hyperscale AI data center project. The contract is valued at approximately 181.2 billion won.

The project involves the construction of large-scale AI data centers across North America, along with power generation and transmission infrastructure needed to support their operations. Under the agreement, LS Electric will supply 345kV transformers to nine substations on a phased basis through 2030.

The rapid expansion of AI computing infrastructure is driving demand for power equipment across North America. As hyperscale data centers increasingly operate at gigawatt-scale capacities, investments are expanding beyond data center distribution systems to include high-voltage transformers, switchgear, circuit breakers and other transmission and distribution equipment.

The latest order also highlights LS Electric’s efforts to expand beyond its traditional strength in power distribution equipment. The company has a portfolio covering the power transmission and distribution chain, ranging from ultra-high-voltage transformers to medium- and low-voltage distribution systems.

Industry observers say securing a 345kV transformer order for a major AI data center project could provide LS Electric with an important reference for pursuing additional business in North America. Large-scale transformer projects typically require extensive manufacturing capabilities, quality assurance and proven delivery performance, making project references an important factor in winning subsequent orders.

LS Electric plans to use the latest project as a foothold for expanding its participation in North America’s AI data center power infrastructure market. The company is targeting opportunities across the entire power system, rather than limiting its focus to individual equipment categories.

NC, jointly led by Kim Taek-jin and Park Byung-moo, said Aion 2 began global early access on Sept. 30 local time.

Players who purchased the game’s Founder’s Pack are eligible to participate in early access and experience the game ahead of its official launch. NC plans to officially launch Aion 2 globally on Oct. 5 following the early access period.

The global service is available through Steam and NC’s PC gaming platform PURPLE. Servers are being operated across four regions — North America, South America, Europe and Asia, including Japan.

The game supports eight languages, including English, Japanese, French, German and Spanish, to accommodate players across its global service regions.

Aion 2 generated strong interest ahead of its official launch. On Sept. 30, the game topped Steam’s global Top Sellers chart, which ranks games based on sales performance.

It also ranked No. 1 in sales in 12 countries, including the United States, France, Germany, Spain and Brazil. The game had surpassed 1.3 million registrations on Steam’s Wishlist before its release.

NC released a launch trailer to mark the start of global early access. The video showcases key features of Aion 2, including free-flight movement, aerial combat and the conflict between the Elyos and Asmodians, along with the game’s PvE and PvP content.

Free flight and aerial combat are among the core gameplay elements carried over and expanded from the original Aion. The sequel builds on the original game’s fantasy world while offering expanded combat and multiplayer content.

NC plans to use the early access period to introduce the game to global players before officially launching the title on Oct. 5.

Further information about Aion 2, including its global service and game content, is available through the game’s official website.

Doosan Enerbility said Wednesday it has signed an approximately 840 billion won ($610 million) contract with Vietnam’s state-owned energy group PetroVietnam (PVN) to build the O Mon 3 combined-cycle gas turbine power plant.

Under the contract, Doosan Enerbility will form a consortium with PTSC, an energy services company and subsidiary of PVN, to carry out the project on an engineering, procurement and construction (EPC) basis. The company will also supply the plant’s key steam turbine equipment.

The project marks the first time Doosan Enerbility’s steam turbine will be installed at a gas-fired power plant in Vietnam. The company will therefore handle both EPC work and the supply of major power-generation equipment, broadening its role in the local power market.

The O Mon 3 plant will be built within the same power complex as the O Mon 4 plant, which Doosan Enerbility won last year. Located about 180 kilometers southwest of Ho Chi Minh City, the plant will have a total generation capacity of 1,155 megawatts and is scheduled for completion in 2030.

The O Mon 3 and O Mon 4 projects are key components of Vietnam’s efforts to expand power generation in its southern region. The projects are expected to contribute to stabilizing electricity supplies as power demand continues to rise.

Doosan Enerbility’s consecutive wins at the O Mon complex are also drawing attention from the industry because they demonstrate the company’s ability to convert an existing project track record into follow-up orders.

The company has worked with different local partners on the two projects, partnering with PECC2 for O Mon 4 and PTSC for O Mon 3. Industry observers say such local partnerships and accumulated project experience could provide a foundation for pursuing additional power-generation projects in Vietnam.

Vietnam is expected to remain a key market for gas-fired power generation as the country seeks to expand its power infrastructure to meet growing electricity demand. Doosan Enerbility plans to leverage its experience from the O Mon 3 and O Mon 4 projects to pursue additional gas-fired power projects not only in the southern O Mon complex but also in northern and central Vietnam.

The latest contract is part of Doosan Enerbility’s broader push to expand its gas-fired power business overseas.

In June, the company won contracts worth approximately 840 billion won for the second phase of the Jafurah cogeneration power plant in Saudi Arabia and approximately 530 billion won for the construction of the Duqm combined-cycle gas power plant in Oman.

In August, it secured another contract worth about 930 billion won for the construction of the Misfah combined-cycle gas power plant in Oman. Last month, it also signed a contract for the construction of the Hadong combined-cycle power plant in South Korea.

Including the O Mon 3 project, Doosan Enerbility has secured five domestic and overseas gas-fired power plant construction projects this year, with a combined value of approximately 3.87 trillion won.

The company has increasingly combined EPC capabilities with the supply of core power-generation equipment in its recent projects. This approach allows Doosan Enerbility to leverage its expertise in gas turbines, steam turbines and other power-generation equipment alongside its plant construction capabilities.

Industry observers say securing large-scale project references in Southeast Asia and the Middle East could strengthen Doosan Enerbility’s position in subsequent bidding opportunities, particularly in markets where electricity demand and investment in power infrastructure are increasing simultaneously.

“By securing the O Mon 3 project following O Mon 4 last year, we have once again demonstrated our competitiveness in Vietnam’s combined-cycle gas power market,” said Lee Sung-yeol, head of Plant EPC Sales at Doosan Enerbility. “PVN is planning additional combined-cycle gas power projects in the southern O Mon complex as well as central and northern Vietnam. Based on the successful execution of the O Mon 3 and O Mon 4 projects, we will actively pursue further orders in Vietnam’s power market.”

Doosan Enerbility plans to use its growing portfolio of gas-fired power projects as a foundation for expanding its presence in Vietnam, the Middle East and other overseas power markets.

According to the Ministry of Trade, Industry and Energy’s September 2026 export and import report released Wednesday, exports reached $120.94 billion last month, up 83.5% from a year earlier. Imports rose 26.0% to $71.09 billion.

The resulting trade surplus stood at $49.85 billion, marking the first time Korea’s monthly trade surplus has surpassed $40 billion.

Exports also set a new cumulative record. Total exports for the January-September period reached $814.5 billion, already exceeding the previous full-year record of $709.3 billion set in 2025.

Average daily exports, adjusted for the number of working days, rose 105% year-on-year to $5.63 billion, surpassing $5 billion for the first time.

September became Korea’s largest month for exports on record, overtaking June’s $102 billion and July’s $99.2 billion.

Semiconductors were the biggest contributor to the record-breaking performance. Semiconductor exports jumped 262.8% year-on-year to $60.3 billion, surpassing the previous monthly record of $46.7 billion set in August.

It was the first time a single export item had generated more than $60 billion in one month. The increase was driven by sustained demand for memory chips, higher export volumes and continued increases in fixed prices.

The strength of semiconductor exports was evident throughout September. Semiconductor exports totaled $16.48 billion during the first 10 days of the month, up 270.1% from a year earlier, accounting for 47.1% of Korea’s total exports during the period. By the first 20 days, semiconductor exports had reached $34.12 billion, up 259.4%, with their share of total exports rising to 47.8%.

The surge reflects the growing demand for memory products amid the expansion of AI data centers and server infrastructure. Demand for high-bandwidth memory (HBM), server DRAM and other high-value memory products has continued to support Korea’s semiconductor exports.

Exports excluding semiconductors also increased 23%, indicating that the September export boom was not limited to the chip industry.

Computer exports surged 435.3% to $7 billion as the spread of agentic AI and rising investment in AI infrastructure boosted demand for servers and related equipment.

Cosmetics exports rose 31.4% to $1.51 billion, reaching an all-time monthly high as overseas demand for Korean beauty products continued to expand.

Ship exports increased 29.9% to $3.76 billion, supported by higher deliveries of liquefied natural gas carriers and very large gas carriers. Petroleum product exports jumped 72.0% to $7.22 billion, partly reflecting higher international oil prices.

Automotive exports, however, declined 5.5% to $6.05 billion. The decline was attributed partly to a base effect stemming from the timing of the Chuseok holiday and weaker used-car exports.

Exports to Korea’s major trading partners also posted sharp increases.

Exports to China surged 122.7% year-on-year to $26.01 billion, exceeding $20 billion for the fourth consecutive month. Exports to the United States rose 137.0% to $24.33 billion.

Exports to the Association of Southeast Asian Nations (ASEAN) jumped 92.6% to $21.29 billion, surpassing $20 billion for the first time. Exports to the European Union increased 20.9% to $8.65 billion, extending their growth streak to 10 consecutive months.

Exports to the Middle East fell 14.1% to $1.6 billion amid difficulties in navigation through the Strait of Hormuz and continued geopolitical uncertainty in the region.

The strong export performance also translated into a sharp increase in Korea’s trade surplus. The cumulative trade surplus for January through September reached $252.5 billion, an increase of $202.7 billion from the same period last year.

The latest figures have strengthened expectations that Korea could reach $1 trillion in annual exports this year. The country would need to maintain strong export momentum during the fourth quarter to reach the milestone.

Kim Jung-kwan, minister of trade, industry and energy, said September exports exceeded $120 billion for the first time, while cumulative exports through September surpassed $800 billion and exceeded last year’s full-year record of $709.3 billion ahead of schedule.

“Annual exports of $1 trillion are now within reach, but the strengthening of global protectionism and tensions in the Middle East remain major variables for Korea’s exports,” Kim said. “We will work to maintain the momentum in exports through the fourth quarter.”

The September figures underscore the growing influence of AI-related demand on Korea’s export structure. While semiconductors remain the main driver, strong growth in computers and other AI infrastructure-related products is broadening the impact of the AI investment cycle across Korea’s export industries.

At the same time, the concentration of export growth in semiconductors and continued geopolitical and trade uncertainties remain key factors to watch as Korea enters the final quarter of the year.

Mercuryo announced in London on September 30, 2026, that it will waive its fees on the first three purchases converting traditional currency into cryptocurrency for new MetaMask users in Canada. The promotion runs from September 30 through October 31, 2026, and covers all available digital tokens in the MetaMask mobile app and browser extension. MetaMask fees and network fees will continue to apply.

Eligible purchases can be made by credit card, debit card or Apple Pay, using either Canadian or US dollars. The offer applies to transactions between C$32 and C$483, according to Mercuryo. The waiver reduces one component of the cost of buying digital assets through the wallet.

“We anticipate strong interest among new MetaMask users in Canada for this offer,” said Arthur Firstov, chief business officer at Mercuryo. “MetaMask is building a new kind of consumer financial experience, and we’re delighted to partner with them to give people more choice, flexibility and control over how they manage their money.” Mercuryo provides payment infrastructure connecting traditional payment methods with digital asset platforms.

Series Order

Part 1 | Is AI Conscious? The Risks of Technology Turning Human Nature into Data
Part 2 | Is AI Conscious? Part II | AI as a Cultural Mirror

By Layne Hartsell

What we see in AI is, in a real sense, us. AI reflects the culture that created it based on enormous amounts of training data. In fact, software, or machines, called “AI” are more precisely described as token-generating machines. Tokens are units of input — letters, punctuation, chunks of data — generated according to probability patterns. Trained on data at a large scale, this software produces output mimicking language as text.

This is not language. Biolinguistics holds that humans have an innate capacity for language, which is generative and creative. We also understand meaning. Basically, human language is not produced by probability matching from external inputs, and neither are intention and consciousness matters of computation. Token-generating machines lack innate faculties — and interestingly enough, the discussion today seems to reverse the mid-20th-century Chomsky–Skinner debate as alluded to above in the first essay of this series. At that time, humans were cast as blank slates; today, machines are treated as having innate faculties. It is an interesting convolution of cognitive and computer science.

Linguist Emily Bender calls AI systems “stochastic parrots” and, importantly, has cataloged harms that have already been done by AI: discriminatory outcomes in housing and criminal justice; denial of healthcare; hate speech along with misinformation and disinformation; deepfake pornography or false compromising photos and stories; and wage theft via algorithmic management. Bender, Alex Hanna and others compare AI to the children’s book Everything in the Whole Wide World, where Grover, a “muppet” from Sesame Street, visits a museum whose categories seem exhaustive but are arbitrary until he reaches the door to “Everything Else” — the outside world. Their point: AI benchmarks claim to measure general “understanding” but are mostly arbitrary — much like benchmarks claiming to measure general “understanding.” Opening out to the actual world is a different matter, that of fullness or integrality.

Artificial intelligence largely reflects the Western culture that produced it — an Echo of Narcissus, with recursion as both a technical and social danger. Technically, problems show up as “model collapse,” degradation of data and runaway feedback loops. Then there are cultural issues such as amplification of the general culture or subcultures. Add the erroneous belief that this software is conscious, and misinformation and disinformation are magnified rather than corrected — making the technology culturally insidious and thus a truly toxic mix, hence dystopia. We look into the mirror, the reflection comes back, more data is generated from that reflection, and the loop deepens into what I’d call “aperspectival psychosis”: a society-wide loss of the ability to distinguish and discern. This is a technological dystopia organized around market mechanics and profit with social isolation, rather than human relationality.

This dystopia shows up with particular force in the technical imposition of symptoms of autism, ADHD and psychosis — what some actually call “brain rot” — because the mind, or intention specifically, is targeted by highly sophisticated digital systems of companies and government. Around twenty years ago, marketers tried to use fMRI to peer directly into brains in order to influence people in the market and society; today, that goal is achieved by hijacking attention and commandeering intention. German-American philosopher Hannah Arendt (1906–1975) warned that when people lack the skills to distinguish accurate information from falsehood, they become easy subjects for totalitarianism. Unfortunately, we are orders of magnitude further into technological dystopia than when Arendt was writing, as algorithms for behavioral modification have become a business model monetizing attention and intention for near “pure profit,” to quote one of the major automotive companies. Yes, they realized that they could collect data from vehicles they sell.

What Computation Is — and Is Not

Computation today means multiple parallel input/output processes, highly complex neural networks, though they are still input/output processes. AI is a Cartesian Theatre of discrete, controllable, algorithmic computation trained on enormous datasets, producing sophisticated token-matching — what Lynn Andrea Stein at MIT calls “a sophisticated symbol-processing abacus.” However, there is no real scientific comparison between AI and embodiment and consciousness, as Nature is clearly “outrageously different,” as quantum physicist Roger Penrose points out. There seems to be no way to reduce consciousness to computation. The issue is that the conflation is dangerous: the belief that AI is conscious permits reduction and objectification in the first place.

Two beliefs circulate consistently: that AI understands language like a person, and that AI explains the human mind or is conscious. Neither is established. John Searle’s Chinese Room thought experiment remains useful: a basic description is that I could be in a room, receive a Chinese character through one door, match it to a character within the room, and then send it out the other door. While I might get the matching correct without error, I still would not understand Chinese, which could be easily tested. There is mimicked language from outside (matching), yet undeniable failure of understanding. Basically, with the machines today, we see data and output, but no theory of mind, that is, if we refer to mind as rooted in embodied contact with the natural world, developed over millennia through culture and the land. Consciousness and intelligence are embodied facts; for humans, they are further embedded in culture, history, emotion, various faculties, and then intention and attention.

Computronium and the Substrate Swap

The ultimate expression of computational reductionism is “computronium,” computer scientists Norman Margolus and Tommaso Toffoli’s term for a hypothetical substance optimized for maximum computational density. It is in the same area as the Matryoshka brain or Dyson swarm (cloud). Yet computronium fails even on materialist grounds. Consider the “substrate swap” thought experiment: scientists replace the neurons in a conscious brain, one by one, with computronium or optimized computational material — let’s call it a crystalline material — performing identical input/output functions, and ask, after each swap, whether the person is aware. If subjective experience vanishes while nothing changes functionally (computation), consciousness is expected to depend on a biological substrate, not computation.

If subjectivity persists, computation might suffice for consciousness. In the latter case, it would presumably be possible to upload into a computer. I have been highly skeptical that consciousness supervenes on computation, and have thought the case does not pass physicality. My view is, as mentioned, that consciousness and mind depend on embodiment and experience or a kind of pragmatic development.

Also, like consciousness, biology is not computational. For example, protein production operates through distributed cellular processes that do not fit into input/output models as was once thought. It is perfectly fine to abstract from nature to produce models, but the gross error is to believe the model. More important than the thought experiments is current anesthesiology research by Stuart Hameroff and Roger Penrose on microtubules, where consciousness is “turned on and off” when the patient goes to sleep. The reasoning in this physical theory of consciousness is that, while consciousness appears not to be computational, it seems to have something to do with quantum processes in microtubules — a biological theory of consciousness.

This “objective reduction” is speculated to be what constitutes conscious events and brings physics, chemistry and biology together. If accomplished, it would be quite amazing and important. While all of the above are interesting and should be studied, it is clear that what looks like a deep mystery to us seems to pose no difficulty for nature itself, which clearly “chooses” in various cases, such as between wave and particle duality, and seems to choose or discern with consciousness in other places — Penrose calls it “proto-consciousness.” If, somehow, it turns out that organic material is absolutely necessary, then the organoids (small brains) that are being produced might be far cheaper and easier to produce, with better computational performance, than silicon-based computation or other computronium.

Perhaps the substrate or brain and computronium are the wrong direction, though they might yield useful information. Perhaps consciousness is an attribute of Nature and we see varying degrees or aspects due to embodiment and experience, dependent on the dependent origination of embodiment and consciousness as in Buddhist philosophy; if so, this would indicate that actual understanding would be beyond our faculties. I am of this thinking.

Mathemythics and the Imitation Game

The Turing Test is a provisional, operational test for determining the ability of a computer to produce textual speech — measuring mimicry, not emotion, imagination, foresight or transcendence. It measures nothing in the realm of embodiment. While Turing thought successfully passing his test would be a milestone, or a canary in the coal mine — a point at which, if achieved, we should give the matter more social attention — he did not think computers were conscious beings. Later, Dutch mathematician Edsger Dijkstra, asking rhetorically, “Do we say submarines swim?”, thought the question uninteresting. Instead of Turing or Dijkstra, many want a mathemythics in AI, something akin to climate economist Steve Keen’s critique of mathematical formalism used to manufacture myths in economics. The patterns are applied widely at times.

What’s happening is a milieu of belief in which computer scientists and CEOs treat the reduction of consciousness to computation as already accomplished — taken up by credentialed people who state as settled fact what remains, according to physicists themselves, a live noncomputational question that apparently involves quantum mechanics. Sycophancy and sensation seem a better explanation for the current AI cultural phenomenon than physics and mathematics.

Historically, modernity’s default was that non-humans lack inner life; with AI, that default flipped. AI is assumed to understand until proven otherwise, a self-sealing loop.

Embodiment and the Social Fabric

British artist and writer John Ruskin observed that human consciousness arose through contact with nature; likewise, American wilderness guide Sigurd Olson recognized, after innumerable canoe trips into the wilderness, that the human psyche is “attuned to woods and fields and waters.” Poetics may provide a bit of help when it comes to the big questions, where the belief that AI possesses consciousness is a categorical mistake, and a mistake magnified into social consequences.

Relational instincts holding communities together — personhood, actual relationship, obligation, care — are threatened whenever mechanization enters our interactions.

When fluent text substitutes for evidence of inner life, criteria for personhood shift from intimacy and mutuality to superficial responsiveness and mindless interaction. However, relationships involve dynamic interaction between people whose interests conflict — that dynamism teaches patience, negotiation, new ideas and ways of being, and tolerance. A system optimized for mechanical operations with relationality stripped out has no such dynamism, and I suspect that skills and habits that have normally sustained reciprocity will atrophy. Therefore, while stupidity will spread, psychological distress would also ensue, since the Self can be defined as the discerning factor of relationality and, when systemically frustrated, will lead to psychological illness.

Researchers such as sociologist Jonathan Haidt are reporting on problems already concerning real relationships: inner life is attributed to the machine, which seems harmless and even “preferable to humans.” When widely believed, it creates structural erosion pushing society towards fragmentation, visible in the global pandemic of loneliness that was already an issue before AI due to social media and market society. The situation is also physically unhealthy: in a 2023 meta-analysis of more than two million people, Chinese epidemiologist Fan Wang found that social isolation and loneliness are each independently linked to higher mortality risk. Insidious and dystopia are appropriate terms to describe the conditions that are worsening.

Part 3 will follow with a discussion of the new technoutopianism and concluding comments.

About the Author

Layne Hartsell, USA (雷恩∙哈特塞尔 / 마이클레인핫셀) — 3E: Energy, Economy, Environment. Research fellow, Center for Science, Technology, and Society, Department of Philosophy, Chulalongkorn University, Bangkok. Research fellow, 3E and Bioregional Letters, North Valley College, Mindanao, Philippines, and the Asia Institute, Tokyo. Past affiliations: Assistant professor, Convergence Studies, Sookmyung Women's University, and the Research Institute for Asian Women / Asia-Pacific Women's Information Network Center, Seoul; Research professor, Sungkyunkwan University and the Advanced Institute of Nanotechnology, Seoul/Suwon; Adjunct lecturer, Mahidol University, Siriraj Medical Center, Department of Molecular Biology and Bioinformatics, Bangkok; Researcher, University of Virginia College of Medicine, Charlottesville.

Korean version available here

KORAIL said on Sept. 30 that it has secured an additional contract to provide procurement consulting services for MRT-7 maintenance parts.

The contract follows an approximately 120 billion won operations and maintenance (O&M) agreement signed with Philippine conglomerate San Miguel Group in April last year.

MRT-7 is an urban railway line being developed by San Miguel in Metro Manila. Under the new contract, KORAIL will support the establishment of a maintenance-parts procurement system needed for the railway’s operations after opening.

Building a Procurement System for the First Three Years

KORAIL will provide consulting services during the first three years of MRT-7 operations, helping San Miguel establish a structured procurement system and strengthen its own sourcing capabilities.

The project will cover the full procurement process, including purchasing plans, technical specification reviews, supplier identification, quotation and contract reviews, pre-shipment inspections, local inspections and technical consulting.

A stable supply of maintenance parts is essential for long-term railway operations, particularly across rolling stock, electrical and mechanical systems.

San Miguel plans to purchase maintenance parts from domestic and international suppliers throughout its 25-year operation of MRT-7. In the second half of this year, it plans to pursue approximately $4 million worth of procurement contracts involving around 20 suppliers.

Opening Opportunities for Korean Rail Suppliers

The project is also expected to provide Korean railway parts manufacturers with a direct route into the Philippine rail market.

KORAIL plans to identify qualified Korean suppliers and connect them with San Miguel, supporting their participation in the MRT-7 maintenance supply chain.

In July, KORAIL held an MRT-7 maintenance procurement briefing attended by more than 30 Korean railway parts companies. The event included one-on-one business meetings between Korean suppliers and San Miguel executives.

KORAIL will continue to support Korean companies throughout the procurement process, including quotation and contract reviews as well as quality inspections of manufactured parts.

Expanding Overseas Rail Business Beyond O&M

The additional contract broadens KORAIL’s overseas railway business beyond traditional operations and maintenance into procurement and supply-chain management.

The model could also create longer-term export opportunities for Korean railway companies. Overseas rail projects generate recurring demand for replacement and maintenance parts over decades of operation, allowing Korean suppliers to participate beyond the initial construction phase.

“This contract expands our involvement in Manila MRT-7 from operations and maintenance into procurement while supporting Korean railway companies in entering overseas markets,” KORAIL President Kim Tae-seung said. “We will use our expertise in railway operations and maintenance to create more export opportunities and broaden overseas sales channels for Korean companies.”

The Korean telecommunications company unveiled its AICC, or AI Contact Center, strategy at a press briefing held at the LG Gasan Digital Center in Seoul on Sept. 30.

LG Uplus is applying AI across the entire customer service process, from AI-powered voicebots and chatbots to its AI Consultation Advisor for human agents and AI Auto QA for automated service quality assessment.

Based on the technology and operational experience accumulated through its own contact centers, the company plans to expand the solutions to enterprise customers and grow its AICC business revenue by more than 50% annually through 2030.

In-House Contact Centers Serve as AI Testbeds

A key element of LG Uplus’ strategy is its in-house approach to AICC development.

The company operates 17 customer service centers nationwide, handling tens of millions of consultations each year. These centers serve as real-world testbeds where new AI technologies can be evaluated using actual customer interactions and operational data before being offered to business customers.

Rather than simply developing and selling AI software, LG Uplus tests its technologies internally, measures customer responses and service quality, and feeds the findings back into product development.

Its customer service specialists can also visit enterprise clients to analyze their operations and design customized AICC systems. Feedback gathered from those deployments is then used to further improve the solutions.

AI is also being incorporated into the development process itself. LG Uplus uses LG AI Research’s EXAONE as well as global AI models to generate consultation scenarios, workflows and AI agents based on a customer’s business objectives and operating guidelines.

The company then tests the systems through simulations designed to resemble real customer service environments before deployment. Once deployed, operational data can be analyzed to identify additional problems and areas for improvement.

AI Handles 8 Million Customer Inquiries Annually

LG Uplus is also dividing customer service tasks between AI and human agents according to the complexity of each inquiry.

Standardized tasks, such as checking mobile plans or subscribing to services, can be handled by AI voicebots and chatbots. About 8 million customer inquiries are currently processed through the company’s AI voicebots and chatbots each year.

Its AI voicebot is designed to understand the context of an entire conversation rather than simply reacting to individual keywords. It can wait when a customer pauses while speaking and use information provided throughout the conversation to determine the customer’s intent.

When a human agent is needed, the AI Consultation Advisor provides real-time assistance.

Conversations between customers and agents are converted into text through speech-to-text technology, allowing AI to analyze the interaction and recommend relevant information, policies and responses.

This reduces the time agents spend searching multiple internal systems and allows them to focus more on customers’ individual circumstances and needs. It can also help less-experienced agents provide more consistent responses.

After each consultation, AI Auto QA analyzes the interaction and identifies individual agents’ strengths and areas for improvement. The system enables a larger share of customer interactions to be assessed under consistent criteria, while allowing service coaches to spend more time on personalized training.

Average Agent Wait Time Falls from 47 Seconds to Five

LG Uplus said the deployment of its in-house AICC has already produced measurable improvements in customer service operations.

The average wait time to connect with a human agent has fallen by about 90%, from 47 seconds in 2024 before the system was introduced to five seconds this year. Average consultation time has also declined by 16% over the same period.

The company is using some of the capacity created by AI automation to strengthen services that require more human attention.

Customers aged 65 and older can receive priority assistance from human agents, while customers with hearing impairments can be connected to sign-language support. LG Uplus also maintains a policy of allowing customers to choose their preferred consultation method rather than requiring them to use AI.

The approach is designed to let AI handle repetitive and standardized processes while human agents concentrate on interactions that require judgment, explanation and personalized support.

Moving Toward an AI-Native Contact Center by 2030

LG Uplus plans to take the model further by shifting from AI-assisted customer service to an operating structure built around AI.

Under its planned AI-Native Contact Center, AI would analyze information such as a customer’s previous consultations and subscription history to better understand the reason for an inquiry and provide relevant information before the customer has to wait for assistance.

The company also plans to analyze customer service data in real time and continuously feed the findings back into the system, creating a cycle in which AI improves as more interactions are processed.

LG Uplus intends to apply the experience gained from its own contact centers to its B2B business, positioning operational expertise alongside AI technology as a key differentiator in the enterprise AICC market.

“In the AI era, competitiveness in AICC depends not simply on the technology itself, but on how quickly customer responses can be translated into service improvements,” said Ahn Hyung-kyun, head of LG Uplus’ Enterprise AI Business Group.

“Based on the experience accumulated through our own customer centers and our in-house capabilities, we will support enterprise customers in their AI transformation and develop AICC into a key growth pillar of our enterprise AX business by 2030,” he said.

The company said it will invest 11.6 billion won ($8 million) to build a samarium metal production facility at its LSCV manufacturing site in Ho Chi Minh City, Vietnam. Trial production is scheduled to begin in December.

The facility will have an annual production capacity of about 240 tons of samarium metal, enough to support the production of approximately 700 to 1,000 tons of samarium-cobalt, or SmCo, permanent magnets per year.

Samarium Gains Importance in Defense Supply Chains

Samarium is a key raw material used in SmCo permanent magnets, which can maintain stable magnetic performance at high temperatures. Their heat resistance and reliability make them suitable for demanding aerospace and defense applications, including fighter aircraft and missiles.

The investment comes as the United States moves to tighten restrictions on Chinese materials used in defense procurement.

From January 2027, U.S. restrictions covering SmCo magnets for Department of Defense procurement are set to extend further across the supply chain. The regulatory shift is increasing demand for alternative sources of samarium and related materials outside China.

LS Eco Energy aims to position its Vietnamese facility as an early production base capable of responding to this shift.

Lynas, LS Eco Energy and LS Cable & System Link Supply Chain

The project also marks the first move into commercial production under LS Eco Energy’s partnership with Australian rare earth producer Lynas.

The two companies strengthened their partnership in July by mutually acquiring convertible bonds worth 30 billion won each.

Under the planned supply chain, Lynas will provide rare earth feedstock, which LS Eco Energy will process into metal in Vietnam. The material will then be supplied to LS Cable & System’s permanent magnet operations.

The arrangement is designed to connect rare earth feedstock, metal processing and permanent magnet manufacturing through a supply chain outside China, linking operations in Australia, Vietnam and South Korea.

For LS Eco Energy, the project also represents an expansion beyond its established power cable business into higher-value strategic materials. The company plans to leverage its existing manufacturing infrastructure in Vietnam as it develops the new business.

Rare Earth Portfolio to Expand Beyond Samarium

LS Eco Energy plans to broaden its rare earth metal portfolio after establishing samarium production.

Future products are expected to include dysprosium, terbium and neodymium-praseodymium, or NdPr, which are widely used in high-performance permanent magnets for electric vehicles, robotics, wind power and other advanced industries.

The strategy is intended to combine LS Eco Energy’s production capabilities in Vietnam with the broader LS Group’s materials and component businesses as global manufacturers seek more diversified sources of critical minerals.

“Even South Korea, which has limited rare earth resources, can become a key player in the global supply chain through the production of high-value-added materials and components,” LS Eco Energy CEO Lee Sang-ho said.

LS Eco Energy plans to begin trial production of samarium metal in December and gradually expand its production portfolio as it builds a broader non-China rare earth supply chain.

Launched in 2004 under the theme of “creating the new based on the old,” Changsinje is an annual traditional music festival organized by a private company. The event is based on the concept of “Beopgochangsin,” which emphasizes preserving traditional heritage while creating new artistic interpretations.

Now in its 21st year, Changsinje has developed into one of Korea’s largest privately organized traditional music performances, with Crown Haetae continuing its efforts to preserve traditional musical forms while presenting them to contemporary audiences.

This year’s festival will center on Sujecheon, a representative piece of Korean traditional court music. Originating from the Baekje-era song “Jeongeupsa,” Sujecheon evolved through the Goryeo and Joseon periods and became an important part of Korean royal court music.

The performance will open with the Seoul Metropolitan Traditional Orchestra performing Sujecheon in a four-part arrangement, accompanied by a choir of 100 Crown Haetae employees.

The program will then feature Ilmu, a ceremonial court dance performed to “Jeonpyehuimun” from Jongmyo Jeryeak, followed by Jeongga Gumeum, featuring male and female vocalists.

A newly created dance performance titled “Obangjeongjae” will also make its debut at this year’s festival. The performance brings together three traditional court dances — Cheoyongmu, Ilmu and Chunaengjeon — into a single choreographic sequence.

The program will continue with the samulnori performance “Dongrak” before concluding with a performance of the original Sujecheon by the Yangju Pungryuakhoe, an ensemble consisting of leading traditional music masters.

The structure of the program combines contemporary interpretations of traditional music with an original performance of Sujecheon, highlighting Changsinje’s focus on both preservation and creative development of Korea’s traditional arts.

Young traditional music talents will also take part in this year’s festival.

The “Young Talents Haneumhoe Changsinje Special Performance” will be held at the same venue at 1 p.m. on October 11. Young musicians supported by Crown Haetae will showcase their skills, developed through weekly training sessions at the Crown Haetae Hall at Namsan Gugakdang in Seoul.

Through Changsinje, Crown Haetae has expanded its cultural activities beyond hosting performances to supporting the next generation of traditional Korean music artists.

A Crown Haetae official said, “Thanks to the continued love and interest of our customers, we will once again present the profound resonance of Sujecheon on one of Korea’s largest traditional music stages.”

“We will continue our efforts to preserve and develop the beauty of traditional Korean music so that more people can enjoy it,” the official added.

The company showcased its next-generation AI infrastructure solutions at Data Centre World Asia 2026, held in Singapore on September 29-30. Key offerings included its Pre-fabricated Modular Datacenter (PMDC) and an AI Factory Full-Stack Service.

The move comes as AI data center operators face growing challenges in securing power, managing heat generated by high-performance GPUs and building infrastructure quickly enough to keep pace with demand.

PMDC is designed to address these challenges by standardizing GPUs, servers, power systems and cooling equipment into prefabricated modules that can be manufactured in advance and installed at the data center site.

Compared with conventional data center construction, in which mechanical, electrical and plumbing (MEP) systems and IT infrastructure are installed largely on site after the building is completed, the modular approach can reduce on-site work and shorten deployment time.

The modular architecture also allows customers to expand capacity as AI computing demand increases. Instead of building a large-scale facility upfront, operators can add modules incrementally based on actual demand.

Industry trends suggest that competition in the AI data center market is increasingly shifting beyond computing hardware toward infrastructure capabilities such as power availability, cooling efficiency, deployment speed and scalability.

The rapid adoption of high-performance GPUs has significantly increased power density and heat generation inside data centers, making power and thermal management critical considerations from the initial design stage.

LG CNS is incorporating capabilities from across the LG group into its PMDC offering. LG CNS provides data center design, build and operation capabilities, while LG Electronics contributes modular cooling infrastructure and LG Energy Solution provides battery solutions.

The company plans to use PMDC as a foundation for a modular AI Factory business. Under this model, customers can deploy container-based AI data centers without having to secure large sites or construct dedicated buildings. Additional containers can then be added as computing requirements increase.

LG CNS has introduced a 4-megawatt IT load module based on NVIDIA's next-generation Vera Rubin GPU architecture and a 1.2-megawatt module based on NVIDIA's GB300 platform. The company plans to expand its modular AI Factory lineup with different configurations to meet varying customer requirements.

LG CNS is also expanding its AI infrastructure business beyond physical facilities.

At Data Centre World Asia, the company introduced an AI Factory Full-Stack Service based on NVIDIA's next-generation Vera Rubin GPUs. The service combines GPU-as-a-Service (GPUaaS), intelligence services supporting real-time training and inference, and power and cooling technologies.

The strategy reflects a broader shift in the AI infrastructure market toward integrated solutions that combine computing resources, data center infrastructure and AI services rather than providing each component separately.

Cooling technology is becoming particularly important as GPU density increases. LG CNS has been strengthening its liquid-cooling capabilities for high-density AI computing environments, including direct-to-chip cooling infrastructure.

The company is also working with NAVER Cloud to build liquid-cooling infrastructure at a data center in Samsong, Goyang, South Korea. The facility is designed to support NVIDIA's next-generation Vera Rubin GPUs, with a total power capacity of 80 MW and the ability to accommodate high-density racks exceeding 200 kW per rack.

LG CNS is leveraging more than four decades of experience in data center design, construction and operation to expand its AI data center business.

The company has been upgrading infrastructure at existing domestic data centers for AI applications while strengthening its capabilities in high-performance GPU infrastructure. It is also pursuing overseas opportunities, including projects in Southeast Asia, North America and Europe.

The company's strategy is to move beyond project-based data center construction by commercializing standardized modules and combining GPUs, power, cooling and operational services. This could allow LG CNS to participate in a wider range of AI infrastructure projects, from initial deployment to subsequent capacity expansion.

Kim Tae-hoon, executive vice president and head of LG CNS's AI Cloud Business Division, said, “AI Factories are a total service that combines data centers, GPU infrastructure and AI services.”

“Customers will no longer want individual services. They will need integrated AI Factories that enable them to generate tangible business results through AI,” he added.

LG CNS said it will continue strengthening its AI infrastructure capabilities and expand its AI Factory ecosystem as it seeks to grow its global business.

The Korean medical AI company recently received approval from Japan’s Pharmaceuticals and Medical Devices Agency (PMDA) for its proprietary lung nodule detection solution, AVIEW Lung Nodule CAD. It is now expanding its Japanese portfolio with AVIEW CAC, an AI solution for coronary artery calcium analysis.

Coreline Soft has signed a supply agreement with M3 AI for AVIEW CAC, which automatically detects and quantifies coronary artery calcification from chest CT images to provide information for cardiovascular risk assessment.

The technology is closely aligned with the company’s opportunistic screening strategy, which seeks to extract additional clinically relevant information from medical images that have already been acquired for another purpose.

A chest CT performed primarily to examine the lungs, for example, can also contain information related to emphysema, interstitial lung disease and coronary artery calcification. AI can analyze and structure these findings without requiring an additional CT scan.

Building on a Network of Around 80 Japanese Hospitals

The AVIEW CAC agreement represents more than the addition of another product to Coreline Soft’s Japanese portfolio.

The company acquired assets related to VUNO Med-LungCT last year, securing an existing customer and distribution base in Japan. It has since expanded its local customer network to around 80 medical institutions while building sales channels with partners including M3 AI and Konica Minolta.

Coreline Soft is now moving from operating acquired products toward introducing its own AVIEW portfolio into that established network.

The recent PMDA approval of AVIEW Lung Nodule CAD provides a regulatory foundation for the company’s proprietary technology in Japan. The addition of AVIEW CAC further broadens its business from lung nodule detection toward comprehensive chest CT analysis.

M3 AI supports the commercialization and distribution of medical AI products by combining access to M3’s extensive physician network in Japan with medical imaging IT capabilities. The partnership could provide Coreline Soft with a platform to introduce additional AVIEW products to Japanese medical institutions over time.

Japan’s Extensive CT Infrastructure Creates Opportunity

Japan represents a particularly relevant market for medical imaging AI because of its extensive installed base of CT scanners.

The country has one of the highest numbers of CT scanners per capita among OECD members, providing a large volume of existing medical imaging data to which AI technologies can potentially be applied without requiring new imaging infrastructure.

Changes in lung cancer screening policy could further expand opportunities. Japan has been moving toward wider use of low-dose CT, or LDCT, for high-risk smokers, with demonstration projects underway in selected municipalities in 2026.

Cardiovascular disease also represents a significant health burden in Japan. Heart disease is among the country’s leading causes of death, making the ability to identify coronary artery calcification from existing chest CT scans potentially relevant for cardiovascular risk assessment.

The combination of extensive CT utilization and an aging population creates a market environment where extracting additional clinical information from scans already being performed could become increasingly valuable.

From Lung Cancer AI to a Chest CT Platform

Competition in medical imaging AI is also evolving beyond the accuracy of individual disease-detection algorithms.

Chest CT images can contain information related to multiple conditions and risk factors, including lung nodules, emphysema, interstitial lung disease and coronary artery calcification. This creates an opportunity for AI companies to move from selling separate disease-specific applications toward providing multiple analytical functions within the same imaging workflow.

Coreline Soft describes this approach as “One CT, Multi-Disease.”

The strategy is designed to expand the role of AI from detecting lung nodules to extracting a broader range of clinical information from a single chest CT examination.

With regulatory approval for its proprietary lung nodule solution and the rollout of AVIEW CAC, the company is beginning to apply that strategy more broadly in Japan.

“Following the recent PMDA approval of our lung nodule detection solution, which established a regulatory foundation for our proprietary product in Japan, we are now expanding our Japanese portfolio with AVIEW CAC,” Coreline Soft CEO Kim Jin-kook said.

“Japan is one of the world’s most active markets for CT utilization, and we plan to gradually expand our strategy of providing diverse clinical information, ranging from lung disease to cardiovascular risk, from a single chest CT scan,” he added.

The investment is drawing attention from the financial industry as a potential strategic move that could give Samsung Electronics more direct exposure to the growing demand for AI data centers.

KB Securities said the investment goes beyond a financial stake, describing it as an opportunity for Samsung Electronics to move closer to the top end of AI data center demand.

Kim Dong-won, head of research at KB Securities, and analyst Lee Chang-min said the investment is significant because Samsung Electronics is gaining direct exposure to the rapidly expanding AI infrastructure market rather than simply investing in an infrastructure company.

Samsung Electronics and five other Samsung affiliates recently announced plans to invest a combined $1 billion in Helix. Samsung Electronics will contribute $500 million, while the remaining five affiliates will invest a combined $500 million.

Helix is an AI infrastructure company led by former Amazon Web Services (AWS) CEO Adam Selipsky. The company is developing an integrated infrastructure business covering data center development and operations as well as power and network infrastructure.

The investment comes as AI infrastructure spending is expanding beyond GPUs and high-performance semiconductors to include data centers, electricity, cooling systems, batteries and communications networks.

Power availability has emerged as one of the key constraints on the expansion of AI data centers. Helix is pursuing a business model that combines data center development with power infrastructure, potentially allowing Samsung affiliates to participate across multiple stages of AI infrastructure projects.

Samsung Electronics can contribute its expertise in AI semiconductors and cooling technologies, while Samsung C&T can provide construction and engineering capabilities for data centers and power infrastructure. Samsung SDS has data center construction and operation capabilities as well as a GPU-as-a-Service business, while Samsung SDI has battery and uninterruptible power supply (UPS) technologies for data centers.

This structure could allow Samsung affiliates to combine their respective capabilities in a single AI infrastructure project.

The investment also comes against a favorable backdrop for the memory semiconductor business.

KB Securities expects DRAM average selling prices (ASPs) to rise by 15% to 20% sequentially in both the third and fourth quarters of this year. The brokerage expects memory prices to continue rising by double digits in the first quarter of next year.

The brokerage also estimates that production capacity from the three major global DRAM manufacturers — Samsung Electronics, SK hynix and Micron — is effectively sold out through 2028.

As AI servers and data centers continue to expand, demand for high-performance memory is increasing structurally, while adding new production capacity requires significant time and investment. This could shift pricing power in the memory market toward suppliers rather than customers, according to KB Securities.

The AI investment cycle is increasingly creating simultaneous demand for data center infrastructure and semiconductors. As hyperscalers expand AI computing capacity, demand is spreading across memory, power systems, cooling, networking and data center facilities.

Against this backdrop, Samsung Electronics' Helix investment could expand its exposure to the AI investment cycle beyond its traditional role as a semiconductor supplier.

The key question will be whether the investment eventually leads to actual data center projects and expanded supplies of Samsung products and technologies, including semiconductors, cooling solutions and other infrastructure components.

KB Securities said, "The scarcity and strategic importance of memory are increasing simultaneously as memory becomes a core asset of AI infrastructure."

The brokerage added that the strategic value of Samsung Electronics' memory business could also be subject to a revaluation as AI infrastructure investment continues to expand.

The report, titled “From Cloud to Edge: The Value of Distributed AI in South Korea,” examines how distributed AI could help the country address structural challenges including an aging population, labor shortages and its high dependence on exports.

Distributed AI combines cloud computing with edge, on-device and on-premise processing, allowing AI workloads to be handled where they are most efficient depending on latency, privacy, bandwidth, energy use and cost.

The approach could be particularly relevant for South Korea, which has strong semiconductor, electronics and telecommunications industries as well as advanced digital infrastructure.

AI Computing Moves Beyond the Data Center

As AI moves into the physical world through robots, vehicles and industrial equipment, computing requirements are changing.

Access Partnership estimates that South Korea’s computing demand for AI inference could increase roughly 30-fold by 2030. Meeting that demand primarily through centralized cloud infrastructure would require data center capacity to expand by around five times.

Under current investment trends, however, the report estimates that available capacity could fall about 86% short of AI inference computing demand by 2030.

Distributed AI offers an alternative architecture rather than a replacement for the cloud. Large-scale workloads can remain in data centers, while applications requiring immediate responses can run on edge infrastructure or directly on devices.

This can reduce network dependence while improving response times, privacy and energy efficiency — factors that become increasingly important as AI is deployed in real-world environments.

For South Korea, the transition could create opportunities across industries in which the country already has significant manufacturing capabilities, including semiconductors, smartphones, automobiles, appliances and robotics.

Autonomous Industry Could Generate $23 Billion Annually

Manufacturing represents the largest economic opportunity identified in the report.

Access Partnership estimates that autonomous industrial operations powered by distributed AI could generate more than $23 billion in annual economic value by 2035.

Industrial robots, autonomous inspection systems and predictive maintenance could use locally processed data to detect equipment problems more quickly, reduce production downtime and improve workplace safety.

The technology could also help address labor shortages as South Korea’s working-age population declines. According to the report, 61% of the country’s manufacturing small and medium-sized enterprises faced difficulties recruiting workers in 2025.

AI and robotics could increasingly support not only repetitive manufacturing processes but also inspection, maintenance and logistics operations.

AI-Enabled Devices Could Strengthen Korean Exports

Exports represent another major opportunity.

The report estimates that distributed AI could generate more than $21 billion in annual economic value by 2035 by strengthening the competitiveness of Korean products in global markets.

On-device AI in smartphones is one example, but the opportunity extends to smart appliances, home and service robots, automobiles and extended-reality productivity devices.

Adding AI capabilities directly to products could increase their value while creating new product categories and strengthening market competitiveness.

The trend could also benefit Korea’s semiconductor industry. As more AI processing takes place on devices, demand could increase for high-performance memory, low-power processors, sensors and connectivity technologies.

Distributed AI therefore has the potential to connect Korea’s existing hardware manufacturing strengths with the next phase of global AI adoption.

Assistive Robots Offer Opportunity in Aging Society

South Korea’s rapidly aging population creates another potential market.

Access Partnership estimates that companion and assistive robotics could generate more than $5.9 billion in annual economic value by 2035.

Mobility assistance robots, fall detection systems, safety monitoring and conversational companion robots could help older people live more independently while reducing caregiving costs and the burden placed on family members.

Edge and on-device AI could be particularly valuable in this area because sensitive personal and behavioral data can be processed locally rather than continuously transmitted to external data centers.

With strengths in robotics, semiconductors and communications infrastructure, South Korea could serve as an important market for deploying and commercializing such technologies.

Physical AI Shifts Focus to Where Computing Happens

As AI adoption expands from software into physical systems, competition is increasingly moving beyond the development of larger AI models.

Where data is processed — and how computing tasks are divided between the cloud, edge infrastructure and individual devices — is becoming a key part of AI performance and deployment.

Access Partnership said South Korea will need approximately $18 billion in additional investment over the next decade, along with stronger infrastructure and global interoperability, to fully capture the economic potential of distributed AI.

For Korea, the opportunity lies in connecting its semiconductor and device manufacturing capabilities with its advanced digital infrastructure. Distributed AI could provide that bridge, extending AI from centralized data centers into factories, vehicles, consumer electronics and robots as the physical AI market develops.

Genesis BBQ Group said on Sept. 30 that BBQ placed eighth in YouGov’s “U.S. Emerging Dining Brands 2026” ranking. BBQ was the only Korean foodservice brand to make the top 10.

The survey, conducted from July 15 to Aug. 2, evaluated emerging brands across the U.S. foodservice industry, including restaurants, quick-service restaurants (QSRs), fast-casual concepts, bakeries and beverage brands.

YouGov first measured nationwide awareness of 70 emerging dining brands among U.S. consumers. The 34 brands recognized by at least 10% of American adults were then evaluated based on five indicators: Net Buzz, Consideration, Net Recommendation, On the Rise and Innovation.

BBQ ranked eighth among the 34 brands based on its overall Emerging Momentum Score. The brand performed particularly well in the Innovation category, ranking third overall.

The Innovation score measures the share of consumers familiar with a brand who consider it to be bringing new ideas to the U.S. dining market. YouGov highlighted BBQ as a brand with notable innovation and said its innovation perception among aware U.S. consumers was higher than that of some global foodservice brands originating in Southeast Asia.

Industry observers say the results point to the expanding visibility of Korean fried chicken in the U.S. dining market. As American consumers show growing interest in new flavors and international food concepts, Korean chicken brands are seeking to differentiate themselves through distinctive recipes, menus and customer experiences.

BBQ has been expanding its presence in the U.S. through store openings and brand-building activities. The company is also seeking to strengthen its recognition among younger consumers. According to YouGov, BBQ’s awareness among U.S. Gen Z consumers reached 17.5%, compared with 10.5% among U.S. adults overall.

The survey results suggest that BBQ’s challenge going forward will be to convert growing brand awareness into sustained customer engagement and store-level growth in the highly competitive U.S. dining market.

A Genesis BBQ Group official said the company will continue to introduce BBQ’s distinctive flavors and brand experience to American consumers as it seeks to strengthen its position as a global foodservice brand.

Dodge AI said on September 29, 2026, that it had raised $2.65 million in a round backed by Accel and Google’s AI Futures Fund. New Build Venture Capital, Antler, Schema Ventures and investors from the SAP ecosystem also participated. The San Francisco company is developing AI tools to investigate failures and manage changes in enterprise applications after they have been installed.

Systems such as SAP, Salesforce and Microsoft Dynamics accumulate company-specific rules and modifications over years of use. When a process fails, IT teams may need to trace the cause through support tickets, software configurations and knowledge held by employees or consultants. Dodge AI says its platform connects those sources to identify likely causes, recommend fixes and document the rules behind each system.

“For a long time, the only way to maintain enterprise systems was to add more people,” said Rebhav Bharadwaj, Dodge AI’s co-founder and chief executive. He said the company aims to help enterprises resolve incidents faster and retain the knowledge gained from maintenance work for future system changes. The platform also supports applications including Kinaxis and Oracle JD Edwards, according to the company.

Dodge AI said it works with more than a dozen enterprises, about half of them publicly listed. In one case described by the company, its platform traced an incorrect warehouse document across SAP, Kinaxis and internal software and delivered a fix within minutes. In another, Dodge AI said it made a customer’s inventory-planning process 132 times faster and reduced order allocation time by eight hours. The customers were not named, and the results were not independently verified.

The company plans to use the operational rules it documents to help AI agents work within production systems. Prayank Swaroop, an investor at Accel, said understanding that context is necessary before agents can operate safely in enterprise applications.

The group said it held talks on Sept. 29 with Yuval Rooz, co-founder and CEO of Digital Asset, and Veep Divakar, head of the Canton Foundation, at Shinhan Financial Group’s headquarters in Seoul. The discussions focused on group-wide cooperation in tokenized finance and digital asset infrastructure.

Digital Asset is the developer of Canton Network, an institutional blockchain infrastructure designed to connect financial systems and assets while maintaining transaction privacy and regulatory compliance. The Canton Foundation supports the governance and operation of the network.

Partnership Expands from Investment to Network Governance

Shinhan Financial Group has been steadily expanding its involvement in the Canton ecosystem this year.

In June, the group signed a memorandum of understanding with the Canton Foundation to participate in network governance. A month later, Shinhan Securities and Shinhan Venture Investment made strategic equity investments in Digital Asset, extending the partnership into the investment sector.

This month, Shinhan Asset Management was approved as a Super Validator on Canton Network, becoming the first Korean financial institution to take on the role.

Super Validators are core participants responsible for supporting network stability and participating in major governance decisions.

The series of moves has broadened Shinhan’s relationship with the Canton ecosystem from an initial partnership to strategic investment and direct participation in network operations.

Connecting Banking, Securities and Asset Management

During the latest meeting, the two sides discussed how Shinhan’s financial businesses could be connected with tokenized financial infrastructure.

Potential areas include payment and settlement infrastructure through the group’s banking business, distribution of tokenized assets through its securities operations, and real-world asset tokenization through asset management.

Such a structure could eventually connect multiple stages of tokenized finance — from asset issuance and distribution to payment and settlement — by drawing on the capabilities of different Shinhan affiliates.

The two sides also explored ways to improve global investor access to Korean financial products through connections with international digital asset infrastructure.

Preparing for Korea’s Tokenized Securities Market

Shinhan Financial Group plans to conduct further technology verification in line with the development of South Korea’s regulatory framework for tokenized securities.

The group also intends to use Canton Network to strengthen connections with overseas tokenized assets and global financial institutions.

As tokenization expands beyond digital asset markets into traditional finance, financial institutions are increasingly examining how blockchain-based assets can be integrated with existing trading, payment and settlement systems.

Shinhan is positioning its collaboration with Digital Asset and the Canton Foundation as part of its broader effort to build the infrastructure needed for that transition.

“As tokenization emerges as a new financial infrastructure in global markets, our collaboration with Digital Asset and the Canton Foundation provides an important foundation for Shinhan Financial Group to proactively respond to changes in the future financial environment,” Shinhan Financial Group Chairman Jin Ok-dong said.

“We will responsibly prepare to expand our connectivity with global digital asset infrastructure in step with regulatory developments and build a service foundation that customers can use safely and conveniently,” he added.

Since entering the remodeling market in 2014, the company has secured projects worth approximately 14.3 trillion won across 46 apartment complexes. Orders secured over the past three years alone totaled about 5.4 trillion won.

The expansion comes as remodeling is increasingly considered an alternative for aging apartment complexes, particularly in areas where high floor-area ratios make full reconstruction more difficult. Unlike conventional redevelopment, remodeling retains much of the existing building structure while upgrading residential space, parking facilities and structural performance.

Construction Experience Builds Remodeling Competitiveness

Apartment remodeling requires different engineering capabilities from new construction because contractors must reinforce and expand existing structures while maintaining their structural integrity.

POSCO E&C has completed three major remodeling projects and is carrying out construction or demolition work at several additional complexes, building experience across different types of urban renewal projects.

The company is also applying technologies including 3D laser scanning and Building Information Modeling to accurately analyze existing structures before construction begins.

For underground parking expansion and structural reinforcement, POSCO E&C has developed and adopted specialized construction methods designed to enable work within the limited space available beneath existing apartment complexes.

These capabilities are becoming increasingly important as more aging residential complexes in Seoul and the surrounding metropolitan area consider remodeling as an alternative to full reconstruction.

Profitability Improves in First Half

The expansion of its remodeling portfolio comes alongside a recovery in earnings.

POSCO E&C posted consolidated revenue of approximately 3.45 trillion won in the first half of 2026, down 6.3% from a year earlier. Operating profit, however, reached 97.4 billion won, reversing an operating loss of 66.9 billion won in the same period last year.

Net income also returned to positive territory at 52.8 billion won, compared with a net loss of 85.1 billion won a year earlier.

The company recorded operating profits in both the first and second quarters, supported in part by improved cost management in its building business. The results indicate that POSCO E&C is placing greater emphasis on profitability and project selection rather than simply expanding revenue.

Digital Technology Expands Role in Urban Renewal

Technology is becoming another important part of the company’s strategy as remodeling projects move from orders into construction.

The use of 3D scanning and BIM allows engineers to create detailed digital models of existing buildings before structural work begins. Such capabilities are particularly relevant to remodeling, where construction must accommodate existing columns, walls and foundations rather than starting from an empty site.

POSCO E&C is also expanding the use of AI, drones and digital construction technologies across its projects to improve site management, construction accuracy and safety.

The company plans to continue targeting major remodeling and urban renewal projects in Seoul and the surrounding metropolitan area while advancing construction methods tailored to aging apartment complexes.

With more of its accumulated remodeling orders moving into the construction stage, POSCO E&C is seeking to translate its 14.3 trillion won project portfolio into a longer-term business base while maintaining the earnings recovery seen in the first half of 2026.

The Korean electronic components maker said Monday that the agreement covers multilayer ceramic capacitors (MLCCs) and inductors for AI servers, with supplies scheduled from Jan. 1 through Dec. 31, 2027.

The deal marks the company’s fifth long-term supply contract disclosed this year, highlighting its push to build a more stable customer base for high-value passive components used in AI infrastructure.

AI servers require significantly greater power than conventional servers, increasing demand for components capable of handling high currents and maintaining stable power delivery under demanding operating conditions.

MLCCs play a key role in stabilizing electrical currents and suppressing signal interference in electronic circuits. In AI servers and data centers, manufacturers increasingly require high-capacitance and highly reliable MLCCs to support increasingly sophisticated power systems.

Inductors are also becoming more important as AI server power consumption rises. They help regulate current fluctuations and reduce noise during power conversion, making high-current and high-efficiency products increasingly important in server power-management systems.

Samsung Electro-Mechanics has been upgrading its MLCC and inductor technologies to address these requirements. The company is focusing on higher capacitance and reliability for MLCCs, while developing inductors with improved current-handling capabilities, efficiency and reliability.

The latest agreement is part of a broader expansion of long-term supply relationships. Samsung Electro-Mechanics currently has long-term MLCC supply agreements with more than 10 global companies and is in discussions with additional customers.

Including the latest deal, the company has secured roughly 3.6 trillion won in long-term supply contracts this year for silicon capacitors, AI server MLCCs and inductors, according to its regulatory disclosures.

Industry watchers say the growing number of long-term agreements reflects a broader shift in the passive-component market as AI infrastructure investment raises both the volume and technical specifications of components required in servers.

The shift is also strategically important for component manufacturers because higher-end AI server products can generate greater value than components used in conventional consumer electronics. Securing long-term contracts can also provide suppliers with greater visibility into future demand while strengthening relationships with major global customers.

Demand for AI server MLCCs is expected to expand rapidly as data center operators increase investments in AI computing infrastructure. According to Goldman Sachs, the global AI server MLCC market is projected to grow from $1.4 billion in 2025 to $5.8 billion in 2030, representing an average annual growth rate of about 34%.

Samsung Electro-Mechanics holds more than 40% of the global AI server MLCC market, according to the company, placing it among the leading suppliers.

As AI servers become more powerful and data centers scale up, the company is expected to continue shifting its component portfolio toward higher-capacity, higher-reliability products while expanding long-term partnerships with global customers.

KB Kookmin Bank said Monday that it signed a strategic memorandum of understanding with BNY on Sept. 28, U.S. Eastern Time, to cooperate on innovation in cross-border digital asset payments and remittance services.

The agreement was signed on the sidelines of SIBOS 2026, the annual global financial services conference organized by SWIFT, at the Miami Beach Convention Center in Florida. Baek Ki-hyun, head of KB Kookmin Bank’s foreign exchange business division, and Jennifer Barker, BNY’s global head of payments and trade finance, attended the signing ceremony along with other officials from the two institutions.

The partnership will initially focus on improving overseas remittance services for foreign customers, including international students. KB Kookmin Bank plans to leverage BNY’s global payments network and “Pay to Wallet” capabilities to provide more convenient and tailored cross-border transfer services.

The two institutions will also explore broader cooperation by combining their global networks and expertise in foreign exchange and payments to improve the overall cross-border transaction experience.

The agreement comes as financial institutions worldwide are expanding efforts to incorporate digital technologies into payment and settlement infrastructure. Industry observers note that partnerships between banks and global financial infrastructure providers are becoming increasingly important as digital assets and new payment technologies gain traction in cross-border financial transactions.

For KB Kookmin Bank, the partnership also provides an opportunity to connect its foreign exchange and payment capabilities with BNY’s extensive global network. The cooperation could help the bank expand services for customers who frequently move funds across borders, particularly foreign residents and international students.

The two sides plan to examine additional opportunities in digital payments, digital assets and next-generation financial infrastructure.

“Through this partnership with a trusted global partner, we have established a foundation for cooperation in overseas payment innovation,” a KB Kookmin Bank official said. “We will continue to lead financial innovation by providing differentiated digital financial services in digital payments, digital assets and next-generation financial infrastructure.”

Shinhan Financial said Monday that Chairman and CEO Jin Ok-dong met with Digital Asset Co-Founder and CEO Yuval Rooz and Canton Foundation President Bib Dhivakar at the group's headquarters in central Seoul to discuss ways to strengthen cooperation in tokenized finance and digital asset infrastructure.

The meeting focused on expanding collaboration between Shinhan Financial and the Canton ecosystem, including the tokenization of financial products and connections between domestic and global financial markets.

Developed by Digital Asset, the Canton Network is designed to connect financial institutions and digital assets while maintaining transaction confidentiality and regulatory compliance. Major global financial institutions, including Goldman Sachs, HSBC, BNP Paribas and the Depository Trust & Clearing Corporation, participate in the network.

Shinhan Financial has been expanding its partnership with the Canton ecosystem throughout this year.

In June, the group signed a memorandum of understanding with the Canton Foundation to participate in the network's governance. In July, Shinhan Securities and Shinhan Venture Investment made strategic equity investments in Digital Asset.

Shinhan Asset Management was also approved this month as a Super Validator of the Canton Network, becoming the first Korean financial company to take on the role. Super Validators support the stable operation of the network and participate in key governance decisions.

Shinhan Financial plans to connect the digital asset capabilities of its affiliates. Discussions covered potential cooperation models linking the banking unit's payment and settlement infrastructure with Shinhan Securities' tokenized asset distribution capabilities and Shinhan Asset Management's real-world asset tokenization business.

The companies also discussed ways to improve global investors' access to Korean financial products by using the Canton Network to connect domestic financial markets with overseas financial institutions and tokenized assets.

Shinhan Financial plans to conduct technology verification in line with the institutionalization of security token offerings in Korea. It also aims to expand connections with overseas financial institutions and tokenized assets through the Canton Network.

“Tokenization is spreading as a new financial infrastructure in global markets, and our cooperation with Digital Asset and the Canton Foundation provides an important foundation for Shinhan Financial to proactively respond to changes in the future financial environment,” Jin said.

“We will responsibly prepare to expand connections with global digital asset infrastructure in line with regulatory developments and build a service foundation that customers can use safely and conveniently,” he added.

The company recently held its “2026 Partners Day,” bringing together key partner companies to share business achievements and discuss future areas of cooperation.

Partners Day serves as a platform for SK hynix and its suppliers to strengthen collaboration in semiconductor technology, manufacturing and quality. As the complexity of advanced memory products such as high-bandwidth memory, or HBM, continues to increase, cooperation with materials, components and equipment suppliers is becoming increasingly important.

Advanced Memory Drives Closer Technology Collaboration

HBM requires more complex manufacturing and packaging processes than conventional memory, including the vertical stacking of multiple DRAM dies and advanced interconnection technologies.

This increases the importance of materials, manufacturing equipment, testing and packaging technologies throughout the production process.

SK hynix is therefore expanding its relationship with suppliers beyond conventional procurement, working with partners from earlier stages of technology development and product validation.

The approach is becoming more important as next-generation HBM and advanced DRAM products require tighter process control, higher performance and increasingly demanding quality standards.

Support Extends from R&D to Mass Production

SK hynix is also expanding programs designed to support partner companies throughout their growth cycle.

Its partnership model covers technology development, testing and validation, mass production and sales, with programs tailored to the different stages of a supplier’s business.

The company plans to provide approximately 1.4 trillion won ($1 billion) in mutual-growth support over the next five years, helping domestic semiconductor materials, components and equipment companies strengthen their technological and business capabilities.

Support includes joint R&D programs, technology evaluation and verification, productivity improvements and financial assistance.

SK hynix is also building the “Trinity Fab” at the Yongin Semiconductor Cluster. Scheduled to open to partners in 2027, the facility will allow suppliers to test and verify materials and equipment under conditions similar to actual semiconductor manufacturing environments.

AI Growth Raises Importance of Supply Chain Competitiveness

The expansion of AI infrastructure is making collaboration across the semiconductor supply chain increasingly important.

As SK hynix broadens its portfolio from HBM4 and next-generation HBM to server DRAM and enterprise SSDs, the competitiveness of the materials, components and equipment companies supporting those products is also becoming a critical part of the overall technology ecosystem.

SK hynix is extending its partnership programs beyond technology and manufacturing. On Oct. 15, the company plans to host “ThanksFULL Day” at Seoul Land in Gwacheon, inviting around 30,000 employees and family members from approximately 130 key partner companies.

SK hynix CEO Kwak Noh-jung and other senior executives are expected to attend the event to recognize the contribution of partner companies.

As AI memory technology advances, SK hynix plans to deepen cooperation with suppliers from technology development through mass production, reinforcing the domestic semiconductor ecosystem while strengthening its competitiveness in the global AI memory market.

The company participated in the TSMC Open Innovation Platform (OIP) Forum held in Santa Clara, California, on Sept. 23, showcasing a broad portfolio of AI memory products ranging from high-bandwidth memory (HBM) to server DRAM and enterprise solid-state drives (eSSDs).

SK hynix has participated in TSMC’s OIP event since 2024. This year, the company presented its technologies under the theme “The Memory behind the Intelligence at Scale,” highlighting the growing role of memory in supporting increasingly powerful AI computing systems.

HBM4 Showcased with NVIDIA Vera Rubin

One of the highlights was SK hynix’s HBM4 and SOCAMM2 products displayed alongside NVIDIA’s next-generation AI superchip platform, Vera Rubin.

The company showcased a 36GB 12-layer HBM4 product and a 96GB SOCAMM2 designed for the Vera Rubin platform, demonstrating memory technologies aimed at supporting higher bandwidth and capacity requirements in next-generation AI systems.

SK hynix also presented an expanded HBM lineup, including a 48GB 12-layer HBM4E, a 48GB 16-layer HBM4 and a 36GB 12-layer HBM3E. A 192GB SOCAMM2 product was also displayed.

The lineup reflects the rapid expansion of AI server memory requirements beyond HBM, as high-capacity and power-efficient system memory becomes increasingly important alongside GPU performance.

AI Portfolio Expands to Server DRAM and eSSD

SK hynix also highlighted products designed for the broader AI data center infrastructure market.

Its server DRAM lineup included a 256GB 3DS RDIMM, a 128GB MRDIMM and a 64GB RDIMM based on the company’s sixth-generation 10-nanometer-class, or 1c, process technology.

For enterprise storage, the company showcased the PEB210 E1.S, which supports liquid cooling, and the PS1101 E3.S, a QLC-based enterprise SSD.

As AI data centers become larger and more power-intensive, memory bandwidth and storage capacity are increasingly being considered alongside energy consumption and thermal management. SK hynix is responding by broadening its portfolio across HBM, server DRAM and enterprise storage.

SK hynix and TSMC Advance HBM5 Collaboration

The partnership between SK hynix and TSMC is also moving into the next generation of HBM technology.

SK hynix received TSMC’s Partner of the Year award in the “3DFabric-HBM Integration” category for the second consecutive year.

The recognition follows collaborative work to validate next-generation HBM5 using TSMC’s CoWoS advanced packaging technology.

As HBM performance increases, advanced packaging is becoming increasingly important in connecting high-bandwidth memory with GPUs and other computing chips. Closer integration between memory and packaging technologies is therefore emerging as a key element in the development of future AI accelerators.

By expanding its portfolio from HBM4 and SOCAMM2 to server DRAM and eSSD while working with TSMC on HBM5 and advanced packaging, SK hynix is strengthening its technology base for the next phase of AI infrastructure development.

Organized by the Tax Free World Association (TFWA), the annual event brings together major companies and industry stakeholders from the global duty-free market. KT&G has participated in the exhibition for five consecutive years to strengthen its presence in the global duty-free tobacco market.

At this year’s event, KT&G showcased its major domestic and international brands, including ESSE, RAISON and BOHEM. The company particularly highlighted ESSE, its ultra-slim cigarette brand, which surpassed KRW 1 trillion in global sales last year. KT&G also used digital screens and displays at its booth to present its global expansion and overseas business performance.

KT&G entered the global duty-free market in 1995 by launching its 88 Lights cigarette in China’s duty-free market. The company currently sells its products at duty-free stores at Korean airports as well as approximately 110 airports and 179 duty-free stores worldwide.

Following its participation in TFWA World 2026, KT&G plans to expand its duty-free business in regions including Europe and Latin America. The company aims to strengthen partnerships with overseas businesses, broaden its duty-free distribution network and enhance the competitiveness of its brands in global markets.

“TFWA World, the world’s largest duty-free exhibition, provided an opportunity to explore new business opportunities with overseas partners and raise awareness of K-tobacco brands,” a KT&G official said. “We will continue to actively expand our global network and strengthen our brand competitiveness in overseas markets.”

KT&G is also broadening its global presence by participating in duty-free and industry exhibitions in countries including Singapore and Germany this year. In August, the company expanded its consumer reach by introducing five ESSE products at the duty-free store of Istanbul Airport in Türkiye, which serves approximately 84 million passengers annually.

Chey attended The Korea Society’s 2026 Annual Gala in New York on Sept. 28 (local time), where Huang was honored with the Van Fleet Award. Chey delivered a celebratory toast recognizing Huang’s leadership in AI and the partnership between Nvidia and SK in the AI semiconductor industry.

The Van Fleet Award is presented by The Korea Society to individuals who have made significant contributions to strengthening Korea-U.S. relations. Chey is also a recipient of the award, following in the footsteps of his late father, Chey Jong-hyun, the former chairman of SK Group.

During his toast, Chey described the partnership between Nvidia and SK in simple terms.

“Nvidia designs the brain, and SK makes the memory,” Chey said. “When Nvidia thinks at lightning speed, SK makes sure that memory is never forgotten.”

He added humorously, “And whenever Jensen forgets something, he knows exactly who to call.”

Chey also highlighted Huang’s emphasis on speed, drawing a parallel with the Korean culture of “ppalli-ppalli,” or “hurry up.”

“Whenever I meet Jensen, his first question is always, ‘Can you make it faster?’” Chey said. “Like Korea’s ‘ppalli-ppalli’ spirit, it seems we share the same mindset.”

Chey connected the history of the Korea-U.S. alliance to the emerging cooperation between the two countries in AI and data centers.

“More than 70 years ago, General Van Fleet built the Korea-U.S. alliance on the battlefield,” Chey said. “Seventy years later, Jensen Huang stands on that front line.”

“But today, that front line is no longer the battlefield. It is the data center,” he added. “And the person leading this advance wears a black leather jacket instead of a military uniform.”

The remarks underscored how the Korea-U.S. partnership is expanding beyond traditional areas such as security and trade into AI, semiconductors and other advanced technologies.

Chey also referred to what he jokingly called “Jensen’s Rule”—“The more you buy, the more you save”—in a lighthearted reference to the massive investments being made in AI infrastructure.

“Yes, our accounting team is still trying to understand that,” Chey said. “But I truly believe in it, because Jensen inspires all of us to think bigger.”

He thanked Huang for inspiring people to “think on a grander scale” and said he was proud to be his partner.

Chey and Huang have developed a close business relationship through their companies’ collaboration in AI semiconductors. SK hynix, a key SK Group affiliate, supplies high-bandwidth memory (HBM) products used with Nvidia’s AI accelerators, making the two companies important partners in the rapidly expanding AI infrastructure ecosystem.

Chey concluded his remarks by asking attendees to raise their glasses to Huang, his family and the Korea-U.S. alliance.

“May our partnership expand much faster than the demand for GPUs,” he said.

Samsung Electronics, Samsung C&T, Samsung SDS, Samsung SDI, Samsung Life Insurance and Samsung Fire & Marine Insurance are participating in the investment. Samsung Electronics will invest $500 million, while the other five affiliates will contribute a combined $500 million.

The investment gives Samsung exposure to the development and operation of AI data centers as demand for computing infrastructure continues to rise with the expansion of generative AI and high-performance computing.

Helix Digital Infrastructure was established by KKR in June to develop and operate hyperscale data centers while integrating power generation, transmission and distribution as well as optical and network infrastructure.

The company is led by Adam Selipsky, former CEO of Amazon Web Services, as co-founder and CEO. Other founding investors include KKR, NVIDIA, Vistra and the Kuwait Investment Authority.

Samsung's participation brings together capabilities spread across its affiliates.

Samsung Electronics can contribute advanced semiconductors, including AI memory, as well as cooling and HVAC technologies. Samsung C&T has construction and engineering capabilities for large-scale data center projects, while Samsung SDS provides expertise in IT infrastructure, cloud services and data center operations.

Samsung SDI can provide battery-based backup power solutions, while Samsung Life Insurance and Samsung Fire & Marine Insurance can support financing and risk-management needs associated with large infrastructure projects.

The investment reflects a broader shift in the AI industry, where data center competitiveness increasingly depends not only on computing chips but also on electricity supply, cooling systems, backup power and network infrastructure.

For Samsung, the Helix investment provides a way to connect its existing businesses across these areas with the growing AI data center market.

Rather than approaching AI infrastructure solely through semiconductor supply, Samsung is seeking to participate across multiple stages of the data center ecosystem, from computing and power to construction, operation and financial services.

The South Korean biopharmaceutical company said Monday it will participate in CPHI 2026, scheduled to take place in Milan, Italy, from Oct. 6 to 8 local time. The event is expected to bring together more than 60,000 industry professionals representing about 2,400 companies from more than 170 countries.

Rather than focusing solely on showcasing its existing products, Celltrion plans to use the three-day event for business discussions with potential partners across several areas.

A key focus will be its novel drug pipeline. Celltrion plans to identify global companies with complementary capabilities in areas including oncology, obesity and metabolic disorders, and autoimmune diseases. The company aims to use such partnerships to expand its next-generation pipeline and accelerate the development of new therapies.

The strategy reflects Celltrion’s broader effort to diversify its growth drivers. While biosimilars remain a core part of its business, the company is seeking additional opportunities in drug development and manufacturing as competition intensifies across the global biopharmaceutical market.

Supply chain resilience will also be a major theme at the exhibition. Celltrion plans to strengthen relationships with existing suppliers of raw materials and contract manufacturing partners while searching for new collaborators to diversify its supply base and improve cost competitiveness.

The company is also looking to capitalize on growing demand for contract development and manufacturing organization (CDMO) services. Celltrion intends to leverage its experience in developing antibody-based biologic drugs and its manufacturing capacity to explore customized partnerships with global pharmaceutical companies.

Industry observers have increasingly focused on the expansion of outsourcing across the biopharmaceutical sector, as drugmakers seek greater flexibility in development and manufacturing. For Celltrion, expanding into CDMO services could provide another avenue for utilizing its accumulated expertise and production infrastructure while creating new business opportunities alongside its biosimilar operations.

Celltrion has participated in CPHI with its own booth for five consecutive years since 2022. This year, the company will operate a roughly 46-pyeong, or about 152-square-meter, standalone exhibition space equipped with dedicated meeting rooms and a lounge.

The company plans to hold a series of one-on-one meetings with potential customers and business partners throughout the event. It will also showcase its research, development and manufacturing capabilities to pharmaceutical and biotechnology professionals from around the world.

“CPHI will provide an opportunity to demonstrate Celltrion’s drug development and manufacturing capabilities on the global stage while further strengthening our global supply chain,” a Celltrion official said. “We will pursue opportunities to strengthen the supply competitiveness of our existing businesses while responding to growing CDMO demand and developing partnerships around next-generation novel drug pipelines.”