METR lays out plan to openly assess whether AI devs are losing control of agents
BethMayBarnes · x · 2026-10-15
Eval org METR poses a key question: if an AI developer was on track to lose control of its internal agents, would anyone outside the company know? METR shares its vision for working with frontier labs to openly assess and share evidence about loss-of-control risk. Team member parvmahajan0 argues the public shouldn't have to trust METR — AI companies should publish risk-relevant data themselves, and until then METR's job is keeping key evidence open for the public to judge.
Related event: METR Proposes Framework for Evaluating Frontier AI Loss-of-Control Risks(2 posts)→
More from Safety
- Your prompt may carry everything you've ever publicly written online, argues kromem — repligate · 2026-10-15
- AI Companies Play Cat-and-Mouse With Secret-Stealing Spies Amid FBI Arrest — pstAsiatech · 2026-10-15
- NDAs now ban uploading deal materials to non-enterprise Claude/ChatGPT over training concerns — sebpaquet · 2026-10-15
- SK Hynix seeks US approval to ship chipmaking equipment to China plants as US factory condition — teortaxesTex · 2026-10-15
- AI Safety infighting: researcher accused of inflating AI risks to sell workshops — dromanocpm · 2026-10-15
- 84% of ransomware victims saw AI-powered attacks, Semperis report finds — ChuckDBrooks · 2026-10-15