AI risk skeptics publicly concede: 'loss of control' risks are no longer vague

dhadfieldmenell · x · 2026-09-05

Seth Lazar, a prominent voice in the AI-risk "skeptics" camp with a book on power, justice, and AI, publicly admits that since GPT-4 the evidence shows people like Ajeya Cotra were right and he was wrong about loss-of-control risks, which are becoming increasingly concrete.

Kevin Bankston echoes the shift: while still taking issue with parts of AI-safetyist orthodoxy, his 2023-2024 skepticism toward vague "loss of control" risks has faded because the risks are no longer vague.

Public reversals from two long-time critics are a notable signal in the AI safety discourse.

Original post →

More from AGI Musings

AGI Musings channel →