Ex-OpenAI researcher: best practices from all AI labs combined would still be sub-Chernobyl

Miles_Brundage · x · 2026-10-04

Former OpenAI researcher Miles Brundage argues AI safety is an industry-wide problem: even combining the very best isolated security and safety-culture practices from every company into one would still be "sub-Chernobyl level." On deaths tied to AI companies, he notes he deliberately used weak language ("tied to," not moral/legal responsibility), but insists that in an industry with billions of users, claiming zero such cases exists means "you're not thinking clearly."

Related event: Ex-OpenAI policy lead Brundage recants "iterative deployment" rhetoric, warns of known-flaw releases(7 posts)→

Original post →

More from AGI Musings

AGI Musings channel →