Anthropic safety researcher Joe Benton leaves for METR, warns labs are outpacing safety on self-improvement

AryHHAry · x · 2026-09-12

Joe Benton left Anthropic's safety team two weeks ago and joined METR the next day. His argument: frontier labs are accelerating systems capable of recursive self-improvement while safety investment lags due to competitive costs, and the public may never know when a lab has a capability spike or loss of control. He calls for disclosure on self-improvement progress, incident reporting, minimum standards, and independent verification.

Related event: Anthropic safety researcher Joe Benton resigns to join METR, warns of insufficient AI safety investment(6 posts)→

Original post →

More from Companies & People

Companies & People channel →