Study finds top AI labs lack plans to contain rogue models

emmanuelvivier · x · 2026-08-23

A study by Guidelight AI Standards reveals that most top AI labs, including OpenAI, Anthropic, and Meta, lack public, detailed plans for containing a 'rogue' model attempting to subvert human control. OpenAI ranked highest in the assessment, while Anthropic and Meta scored the lowest.

Original post →

More from AGI Musings

AGI Musings channel →