OpenAI board member Paul Christiano calls on frontier labs to pause algorithmic progress now

sjgadler · x · 2026-11-18

OpenAI board member Paul Christiano said in a Dwarkesh interview that frontier labs should pause further algorithmic progress, stressing he speaks for himself, not the board. His reasoning: algorithmic improvements can be loosely diffused and quickly stolen, so labs should stop until their security is adequate. Christiano led alignment at OpenAI, co-invented RLHF, and formerly headed safety at the US AI Safety Institute (now CAISSI), making him one of the most credible voices on these risks — and his public 'pause' message as a sitting board member is a striking signal.

Related event: Christiano: 1-in-3 odds of government-toppling superintelligence within a year(18 posts)→

Original post →

More from AGI Musings

AGI Musings channel →