– Focus on building AI systems the public can trust
The Artificial Intelligence Safety Institute (AISI) of the Electronics and Telecommunications Research Institute (ETRI) announced that it held a joint meeting of the AI Safety Consortium on December 17 at the Josun Gravity Seoul Pangyo Hotel, with leading domestic experts in AI safety in attendance.
The meeting served as a platform to review developments in international AI evaluation and safety networks, share AISI’s research outcomes from the past year, and assess progress on key agenda items pursued by the AI Safety Consortium. Discussions focused on both global trends and Korea’s domestic efforts to institutionalize AI safety.
Korea is strengthening its national AI strategy with the stated goal of realizing an “AI Basic Society,” emphasizing the creation of an innovation-driven AI ecosystem, comprehensive AI transformation (AX) across industries, and leadership in global AI governance. Within this framework, ensuring AI safety that citizens can tangibly experience has emerged as a core policy priority and a benchmark of global competitiveness.
In response to these policy demands, AISI has pursued a range of initiatives aimed at strengthening AI safety and trustworthiness. These include evaluating the safety and reliability of domestically developed foundation models and participating in joint assessments through international AI safety networks.
AISI has also worked closely with the AI Safety Consortium—established as a central hub for AI safety research in Korea—to expand research dissemination, share safety evaluation technologies, and facilitate discussions on policy and institutional frameworks. The consortium has functioned as a key collaborative platform linking research, industry, and policy communities.
Key agenda items at the meeting included updates on the activities of the international AI Safety Institute network, a review of AISI’s research achievements over the past year, and reports on progress and future plans across the consortium’s working groups.
Participants were also briefed on outcomes from the second Global AI Safety Network Meeting, held in early December in San Diego, United States. That meeting addressed changes to the network’s structure, including revisions to institute naming conventions, the reorganization of the chair country role—now led by the United Kingdom—and adjustments to governance and operational frameworks.
AISI shared its assessment of these changes and outlined potential response strategies and applications within Korea’s domestic AI safety framework.
During the meeting, AISI publicly presented for the first time a dataset and a draft framework for AI safety evaluation developed over the course of the year. Expert feedback was solicited to improve the framework’s practical applicability and methodological robustness. The institute plans to continue technical reviews and joint research to further advance Korea’s AI safety evaluation system.
Looking ahead, the AI Safety Consortium aims to broaden participation by attracting additional companies and institutions, thereby expanding Korea’s AI safety ecosystem. The consortium also plans to more actively identify and develop policy and research agendas related to AI safety.
As part of this effort, the meeting addressed procedures and roles for new member institutions. Organizations expressing interest may apply through the AISI website, where they are initially registered as associate members before undergoing review for full membership status.
“Myung-Joo Kim, Director of the ETRI Artificial Intelligence Safety Institute,” said that next year is expected to mark a turning point as national AI policies and global governance structures undergo significant change, with competition and cooperation among countries becoming increasingly intertwined.
“ETRI will take a leading role in building a robust AI safety ecosystem,” Kim said. “Through the AI Safety Consortium, we aim to secure both public trust and global competitiveness by ensuring that AI systems are safe, reliable, and aligned with societal needs.”
