Japan AI Safety Institute (Japan AISI / J-AISI)

What it does

Established 14 Feb 2024 by Cabinet decision (referencing the 2024 Integrated Innovation Strategy). Secretariat hosted at the Information-technology Promotion Agency (IPA), a METI-affiliated incorporated administrative agency. Develops AI safety evaluation methods, standards, red-teaming guides and testing environments; coordinates internationally. Explicitly NON-regulatory — it cannot compel, licence or fine. Japan's AI Promotion Act (Act on Promotion of R&D and Utilization of AI-Related Technologies, in full effect Nov 2025) contains no monetary penalties at all; enforcement is reputational/administrative guidance.

Its powers

research

Honest assessment

A real, monitored mailbox at a real agency, which is more than most peers offer. But Japan AISI runs no public consultation or comment docket: its outputs (evaluation guides, red-teaming methodology, data-quality guidebook) are published finished. Realistically a foreigner can write and may get a courteous reply or none; there is no process that obliges anyone to read it.

Notes

Director Akiko Murakami (non-full-time). Advisors include Arisa Ema and Hiroaki Kitano. Site makes no explicit mention of frontier or catastrophic risk — framing is generic 'AI safety'.