KAIST has been selected for Microsoft's global AI safety research program, the only South Korean university chosen.

KAIST said on Monday that Samuel Sohn (손수엘), a professor in its School of Computing, was selected to receive research support under the 'External Red Teaming Alliance (EXTRA)' run by the Microsoft AI Red Team.

EXTRA is a research support program launched this year to work with universities and research institutions worldwide to test AI systems for security vulnerabilities, potential misuse and alignment issues. About 10 universities from six continents are taking part, including KAIST.

KAIST will receive $25,000, about 37 million won, in research funding to study AI safety, security and responsible AI technologies.

Sohn's research team studies attack techniques and defense technologies for generative AI and large language models, including model theft, personal data extraction and prompt injection. It is also developing technologies to improve the safety and reliability of AI agents in real service environments, such as web agents and agentic browsers.

Sohn said, "I will systematically analyse security threats that could arise in real AI service environments and contribute to developing AI technologies that anyone can use with confidence."