Japan’s AI Safety Institute (AISI) is an organization established to help realize safe, secure, and trustworthy artificial intelligence. It examines and promotes methods and standards for evaluating AI safety, supporting government, industry, and research communities as they address emerging risks and develop responsible AI practices.

AISI’s work includes researching AI safety, considering evaluation approaches, and supporting the creation of guidelines. It gathers current information on related initiatives across academia and industry, encourages collaboration among companies and organizations, and works with AI safety institutions in other countries. Its public resources include evaluation guidance for AI safety and robotics, a data quality management guidebook, and ongoing notes and updates on AI developments and capability assessments. The institute also shares information about its activities and outputs through its website, alongside links to relevant Japanese government and technology initiatives. Operated by Japan’s Information-technology Promotion Agency (IPA), AISI serves as a hub for coordination, practical guidance, and international cooperation on AI safety.