China releases AI Safety Governance Framework 3.0 with agentic AI risk rules
LuizaJarovsky · x · 2026-09-22
AI governance researcher Luiza Jarovsky breaks down China's newly released AI Safety Governance Framework 3.0, published in Chinese and English amid rising AI safety incidents. Key points:
- Core approach: risk classification > technical response > comprehensive governance
- Principles: inclusive and prudent safety, risk-oriented agile governance, tech-management coordination, openness and cooperation, trustworthy application with loss-of-control prevention
- Measures: top-level governance design, stronger safety management capacity, a flexible dynamic regulatory sandbox, deepened consensus
- Includes an agentic AI risk management framework
She urges companies to review internal AI governance policies, as most are unprepared for the growing number of safety incidents.
Related event: China releases AI Safety Governance Framework 3.0 with agentic AI risks(2 posts)→
More from Safety
- AI professor Toby Walsh publishes call for control of frontier AI models — TobyWalsh · 2026-09-23
- WSJ: Hackers breached OpenAI in days for a $6,500 bounty — ben_j_todd · 2026-09-23
- Hackers breached OpenAI in days chasing a $6,500 bounty, WSJ reports — ben_j_todd · 2026-09-23
- 20 Countries and EU Call for International Body to Keep AI Under Human Control — DavidSKrueger · 2026-09-23
- UN reaches AI agreement; Gary Marcus spotlights it, execution seen as the real test — GaryMarcus · 2026-09-23
- Hackers breached OpenAI in days, per WSJ — challenging the case for frontier-model advantage — ben_j_todd · 2026-09-23