Glossary
AI Safety Governance
Institutional arrangements aimed specifically at identifying, reducing and overseeing severe AI-related risks.
Last reviewed 2026-08-22
In plain language
The phrase gained currency as governments created safety institutes, evaluation programs and international dialogues on frontier or highly capable models. It overlaps with product safety, biosecurity, cybersecurity and conventional AI regulation. Critics warn that a narrow safety agenda can sideline labor, equality and civic-space harms; proponents argue catastrophic and systemic risks need dedicated institutions.
Why it matters for AI governance
Readers should check whether a “safety governance” proposal includes public accountability and a broad harm taxonomy, or only a subset of technical risks.
Authoritative sources
- Hiroshima AI Process (opens in a new tab) — G7 / Government of Japan