Public research body · Science policy · Pangyo Global R&D Center / Seongnam-si, Gyeonggi-do · Est. 2024
The Korea Artificial Intelligence Safety Institute (Korea AISI; also referred to as the Korea AI Safety Institute) is a South Korean government-linked AI safety research organization housed within the Electronics and Telecommunications Research Institute (ETRI). It was established to serve as a national hub for AI safety, with a mandate to evaluate risks associated with AI models/systems and to develop and research technologies, policies, and assessment approaches intended to prevent or mitigate those risks.
Operationally, the institute is organized into three internal offices aligned to (1) AI safety policy and external cooperation, (2) AI safety assessment (including risk definitions, safety framework development, and evaluation), and (3) AI safety research (technology planning, research, and trends analysis). The institute also positions itself as an international collaboration partner—representing the Republic of Korea in working with AI safety research institutes worldwide and aiming to strengthen standards and international solidarity on AI safety.
No curated programs yet. Check the careers page ↗.
Publications are not yet indexed for this organization. Read them on msit.go.kr ↗.
NAIRL reported a strategic partnership with AISI, including a Letter of Intent signed at AISI headquarters in Pangyo on April 20, 2026, focused on mutual cooperation and advancing Korea’s AI ecosystem for innovation and safety.
MSIT signed an MOU with Anthropic for global AI safety and security cooperation involving AISIMSIT announced an MOU with Anthropic to cooperate on AI safety and cybersecurity, including cooperation with Korea’s AI Safety Institute (AISI) on AI model safety and autonomous agent safety evaluations and other safety-related work contextualized to Korea.
AISI and OpenAI signed an MOU to strengthen cooperation on AI safetyMSIT announced that AISI and OpenAI signed an MOU on June 17 to establish a cooperative framework for safety of frontier AI models in high-risk areas, including knowledge/best-practice sharing on evaluation methodologies and benchmarks and development of an evaluation framework reflecting the Korean language and social context.
AISI announced a Director appointment as Korea’s inaugural AI Safety Institute Director (effective Nov 12, 2024)MSIT announced the appointment of Professor Kim Myuhng-joo as inaugural Director of the AI Safety Institute, authorized by ETRI, with a term from November 12, 2024 to November 11, 2027.
MSIT described the planned establishment and operation model for the AI Safety InstituteMSIT’s English press release outlines the institute’s intended missions and functions, including AI risk definition/evaluation, policy research, AI safety cooperation, and AI safety R&D, and states the institute would be established under ETRI and begin operations in November 2024.