Cambridge prof David Krueger puts AI catastrophe risk above 50%, says everyone is understating it

KatjaGrace · x · 2026-09-10

Cambridge professor David Krueger says "everyone is still understating the risk," putting his personal estimate of AI catastrophe probability above 50%, with his main hope being that we stop building more powerful AI ASAP. He relays a friend at an AI company estimating a 10% chance within the next 4 months absent serious governance intervention.

Quoter KatjaGrace adds that "whole digit % in the coming year" sounds closer to what people actually believe, arguing the message gets chronically understated at every step of relaying.

Original post →

More from AGI Musings

AGI Musings channel →