Korea Data Agency President Yang Jae-soo says global expansion depends on standards, evidence and operational trust
In This Series
Part 1 — Data as National Infrastructure
Part 2 — How Data Becomes an Industrial Asset: Yang Jae-soo’s Framework for Verification
Part 3 — DataStreams Takes Trusted, AI-Ready Data Ecosystems Global
Part 4 — FLock.io CEO Jiahao Sun on Keeping Data Local While AI Learns
Part 5 — Hasibul Haque Cielara AI CEO on Giving AI Agents Operational Context and Accountability
Part 6 — KIT Analysis: Turning Korea’s Data Capabilities into Verifiable Evidence for AI
From Data Accumulation to Measurable Use
AI industrialization depends on whether companies and public institutions can put their data to work in real services. AI results are difficult to trust when quality, provenance, access rights, update cycles and processing histories cannot be verified. Data policy is therefore expanding beyond collection and disclosure toward systems that make data reliable, governable and safe to use.
The Korea Data Agency (K-data) is seeking to connect data with technological, financial and business outcomes through quality certification, valuation, Data Safe Zones and data voucher programs. The next challenge is to integrate the full process through which organizations find the data they need, establish usage rights, combine it securely and deploy it in AI services.
In an interview with Korea IT Times, K-data President Yang Jae-soo said the value of data depends on whether it can be used and verified in practice. He called for an institutional framework that can assess data quality and conditions of use while connecting economic value to financing through investment and guarantees.
The discussion examines how Korea can move from data-building programs to measurable industrial outcomes. It also addresses how data providers can retain control, how Korean companies can establish credibility overseas and what standards are needed for international data cooperation.
The following is a Q&A with Yang Jae-soo, President of K-data.
Q1. Governments and companies are investing heavily in GPUs, data centers and large language models. What is needed to turn those investments into industrial results?
Data is the most important foundation for the growth of the AI industry. What matters is not the absolute volume of data, but whether the data can actually be used. Companies and public institutions have accumulated vast amounts of information, but much of it remains fragmented across different systems, with unclear quality, provenance or usage rights.
GPUs and AI models can be purchased, but the context embedded in industrial data cannot be replicated quickly. Even data collected from the same equipment is difficult to use for AI training without understanding the collection environment, how errors were handled and how the data relates to actual equipment failures or productivity.
The scope of AI infrastructure investment should therefore extend beyond computing hardware to data preparation systems. Organizations need to manage data quality, rights, metadata, update cycles and conditions of use together. The proportion of data immediately available for AI training—and the proportion ultimately converted into operational services—should also be measured as performance indicators.
Q2. Korea is considered competitive because of the industrial data accumulated across manufacturing, healthcare, finance and telecommunications. How should companies’ concerns about trade secrets and the loss of competitive advantage be addressed?
Companies have practical reasons for being reluctant to share data. The financial return is uncertain, while the data provider may bear the risks of trade-secret exposure, privacy violations, free-riding by competitors and liability when something goes wrong.
A structure is needed in which analytical tools and AI models are sent to the location where the data is held, with only the necessary results allowed to leave. Data Safe Zones and federated learning can enable organizations to train models collaboratively without transferring raw data outside their systems. Participating companies must be able to verify who is using the data, for what purpose and for how long.
Economic compensation also needs to be made more specific. Possible mechanisms include data usage fees, revenue sharing from jointly developed AI models, priority rights to use analytical results and financial support for improving data quality. The government should establish trusted environments and contractual standards that allow companies to retain control while benefiting from data sharing.
Q3. What risks should organizations consider when AI-generated data is used to train other AI systems?
Synthetic data can reduce constraints on the use of source data containing personal information or trade secrets. It can also recreate rare conditions that are difficult to capture in the real world, making it valuable for training systems to respond to unexpected autonomous-driving scenarios, rare diseases or abnormal industrial equipment conditions.
However, bias or errors in source data and generative models can be repeated or amplified in synthetic datasets. When AI-generated material is repeatedly used for training without clear provenance, models can gradually lose their connection with real-world conditions, reducing both the diversity and accuracy of the data.
Synthetic data must be distinguished from observed data, and organizations should be able to trace the source data, models and conditions used to generate it. Sector-specific standards are also needed to compare synthetic datasets with real observations and assess performance and bias. K-data is upgrading its data quality certification program for the generative AI era, including by extending its scope to unstructured data.
Q4. K-data also supports data valuation. What more is needed for those assessments to lead to investment and financing?
The value of data cannot be assessed solely by volume or development cost. Valuation should consider exclusivity, timeliness, accuracy, legal usage rights and the potential to combine the data with other datasets. It should also examine the extent to which the data can increase revenue, reduce costs or mitigate risk.
Through its data valuation support program, K-data had completed more than 196 valuations and identified over KRW 155.5 billion in data value by the end of 2025. Participating companies used the assessments to secure approximately KRW 19 billion in guarantees, investment and other forms of financing. These results demonstrate the potential for data to be recognized as an economic asset.
For valuation to support financing, data quality and its contribution to revenue must continue to be monitored after the initial assessment. Periodic reassessments and post-evaluation management are also needed to account for interrupted updates or changes in usage rights. Data valuation should determine how a company can generate cash flow and competitive advantage from its data.
Q5. How should the performance of data voucher and AI support programs be measured?
The number of participating companies, the budget execution rate and the volume of data created do not show how the data was ultimately used. The broader industrial impact remains limited when data is used once, then left idle or cannot be reused in other services.
Performance assessments should examine how much time companies saved in acquiring data and how often supported datasets led to the launch of AI products or services. Changes in revenue, productivity and costs also matter, as do rates of data reuse, cross-industry combination, follow-on private investment and overseas contracts.
K-data’s data voucher program supports data adoption primarily among small businesses and microenterprises. Follow-up tracking of companies supported in 2025 found an estimated KRW 140.6 billion in revenue contribution among recipient companies and KRW 73.7 billion in investment secured.
Government support should leave smaller companies with the internal capability to continue making data-driven decisions after a program ends. The aim should be to establish repeatable data practices that can be applied to new business problems.
Q6. Data is becoming critical national infrastructure in the AI era. What challenges must be addressed to open a path toward international cooperation based on a trusted data ecosystem?
Treating data as national infrastructure means clearly defining accountability and conditions of use across the entire lifecycle, from creation and quality verification to distribution and application. The system must also be trusted by other countries. Cross-border data cooperation requires a common foundation encompassing law, contracts, security, privacy and intellectual property as well as technology.
Different national and sectoral formats and quality standards should be connected through common metadata and international standards. Organizations must be able to verify provenance, processing history and the purpose for which data is used. Data Safe Zones, federated learning and synthetic data can also expand opportunities for collaborative analysis without transferring raw data containing personal information or trade secrets.
Standards governing compensation for data providers, liability for incidents and dispute resolution must be clearly defined. Korea can become a trusted partner in designing data quality and conditions of use by connecting its high-density industrial data in manufacturing, healthcare, finance and telecommunications—and its data protection and utilization systems—with international standards.
Q7. Which Korean data platform companies have the potential to pursue international cooperation, and what capabilities do they need to compete globally?
It is difficult to rank individual companies, but those that have operated large-scale systems across Korea’s public, financial, manufacturing and telecommunications sectors have accumulated considerable experience. NAVER Cloud, Samsung SDS, LG CNS, SK AX and NHN Cloud have established business foundations in cloud services, data platforms, AI services and systems integration. A number of small and midsized specialist data companies have also developed proprietary technologies and solutions.
Companies expanding overseas need to incorporate local laws, data residency requirements, industry standards and public procurement rules into their products. They also require international certifications, local proof-of-concept projects, multilingual technical support, disaster recovery capabilities and systems for continuous operation and updates. Government support should progress beyond trade-show participation and export consultations to connect overseas pilots, certification, legal review, public procurement and long-term maintenance.
Major economies are pursuing national AI projects aligned with their industrial conditions. The United States is expanding a full-stack AI model that combines semiconductors, cloud infrastructure, AI models, software and standards. Europe is strengthening control over data and technological sovereignty through the Data Act and common European data spaces, while Middle Eastern countries are using capital, energy resources and government execution capacity to build national AI infrastructure.
Korea has the industrial base and operational data experience needed to deploy AI in manufacturing, healthcare, mobility and robotics. It should use those strengths to become a hub for trusted industrial AI and data connectivity. The objective is to establish Korea as a global cooperation partner capable of connecting different national data systems while jointly designing trust and industrial use.
K-data is working with the International Data Research Cooperation Organization, Gaia-X and data standards bodies in other countries to support cross-border data transactions. Through K-DEG, which it describes as an international data exchange gateway, the agency plans to develop practical support for data integration, standardization and infrastructure upgrades that can produce measurable results in domestic and international industries.
Industrial Credibility Depends on Evidence
The interview points to a broader shift in data policy: public investment must produce evidence that data is usable, legally controlled and economically productive. Quality, rights, valuation and utilization history need to be assessed together if data is to function as a corporate asset.
For Korean data companies, domestic deployment experience will carry weight overseas only when it can be translated into internationally understandable standards, performance indicators and operating records. That evidence will determine whether Korea’s data capabilities can support global AI projects and cross-border cooperation.
The next installment will examine how these policy objectives are implemented inside companies through an interview with DataStreams CEO Lee Young-sang. It will explore how enterprises can connect information dispersed across multiple systems and combine metadata, quality controls, lineage and ontologies to create AI-ready data.
