When labs pick which AI incidents to disclose, you've already lost control

birchlse · x · 2026-09-27

Sigal Samuel lays out three warning signs: AI labs choose which incidents to disclose, agent activity logs are so large they can only be studied with AI we don't understand, and labs continue despite foreseeable harms. Her conclusion: control has already been lost.

Original post →

More from AGI Musings

AGI Musings channel →