South Korea's Ministry of Science and ICT said on Tuesday that the AI Safety Institute and OpenAI signed a memorandum of understanding to ensure the safety of advanced AI models in high-risk fields.

The signing ceremony was held at the Electronics and Telecommunications Research Institute's Seoul office.

The agreement expands into the safety field the AI cooperation framework built through an MOU signed by the ministry and OpenAI in October last year. It was 추진됐다 based on cooperation measures discussed in two meetings this year between the ministry's second vice minister and senior OpenAI officials.

The AI Safety Institute and OpenAI will share knowledge and best practices related to safety evaluation methodologies and benchmarks by high-risk field. They will also exchange technical information to develop an evaluation system that reflects the Korean language and Korean societal context. They will cooperate to establish an internationally applicable AI safety evaluation system as well.

This is the fourth time OpenAI has signed an MOU with a national AI safety institute, following the United States, Britain and Japan. South Korea will strengthen its role in an international cooperation network that establishes standards for verifying and evaluating AI risks through this agreement.

Lee Jin-soo (이진수), director general for AI Policy Planning at the ministry, said it is time to strengthen cooperation with global leading companies and advance an AI safety evaluation system to ensure the safety of cutting-edge AI models such as high-performance AI and autonomous agent AI. He said he hopes the agreement will help the two organisations cooperate to secure advanced AI safety in response to rapidly changing global AI technology and usage environments.

Kim Myung-joo (김명주), head of the AI Safety Institute, said safety evaluations in high-risk fields are becoming essential as AI's impact on national critical infrastructure and security grows. He said he would contribute, together with OpenAI, to scientifically verifying advanced AI risks and building an evaluation system accepted internationally.

Isanghyeon (이상현), head of OpenAI's Asia-Pacific policy, said South Korea is an important country where the use and innovation of AI technology are occurring rapidly, and the AI Safety Institute plays an important role in international discussions on responsible AI development. He stressed that it is meaningful to share knowledge and best practices on AI safety evaluations in high-risk fields and contribute to developing reliable AI and creating a safe usage environment.

The AI Safety Institute and OpenAI plan to finalise specific cooperation tasks and a schedule after working-level consultations following the agreement.