The Artificial Intelligence Security Institute (AISI) of the Electronics and Telecommunications Research Institute (ETRI), which serves as the control tower for domestic AI safety research, has released its achievements over the past year and has begun to strengthen cooperation in earnest to build an AI safety ecosystem.

On the 17th, ETRI AISI held a joint AI Safety Consortium meeting with domestic AI safety experts at the Chosun Gravity Hotel in Pangyo, Seongnam, and shared the research results of the past year. The meeting focused on key trends in the international AI assessment and safety network and the progress of the consortium's core agenda.

With the goal of realizing an "AI-based society," South Korea is pursuing the creation of an AI innovation ecosystem, a comprehensive AI transition (AX), and leadership in global governance. In this process, ensuring AI safety tangible to the public is emerging as a crucial element and a key criterion for global competitiveness.

AISI has been working to ensure AI safety both domestically and internationally, including through safety and reliability assessments of independent AI foundation models and participation in joint assessments of international AI safety networks. In particular, we have strengthened the foundation for multifaceted cooperation, including disseminating research results, sharing safety assessment technologies, and discussing policies and systems, in cooperation with the AI Safety Consortium.

The meeting presented the current status of the International AI Safety Research Institute Network's activities, AISI's research achievements over the past year, and the results and future plans of each consortium division. Key findings from the 2nd Global AI Safety Network Conference, held in San Diego, USA in December, were also shared. This international conference discussed changes in the international cooperation environment, including changes to the institute's name, a reorganization of the chair (UK), and adjustments to the operational system. AISI presented its response and proposed domestic implementation plans.

At this meeting, AISI unveiled for the first time the dataset and draft safety assessment plan it developed this year. Experts reviewed the data to gather feedback on how to improve usability and effectiveness, and the organization plans to continue technical review and joint research to advance the domestic AI safety technology system.

The AI Safety Consortium plans to expand the domestic AI safety ecosystem by encouraging participation from more companies and organizations and actively explore policy and research agendas. New participating organizations can apply for membership through the AISI website. After registering as associate members, they can become full members after review.

Kim Myeong-ju, head of ETRI's Artificial Intelligence Safety Research Center, said, "Next year will be a turning point where national AI policies and global governance will undergo significant changes, and competition and cooperation between countries will become intricately intertwined." She added, "ETRI will take the lead in building a robust AI safety ecosystem and actively expand the activities of the AI Safety Consortium to simultaneously secure national safety and security and global competitiveness."