Dario urges pacing the frontier; ex-OpenAI safety lead says p(doom) is endogenous

soumitrashukla9 · x · 2026-09-14

Anthropic CEO Dario Amodei published a new essay, "We Must Pace the Frontier," arguing the AI industry should slow down, with a three-part plan. Anthropic unilaterally commits to the first step: granting third-party evaluators permanent, employee-level access to its systems to verify safety measures, report incidents, and assess alignment during training.

Quoting the essay, ex-OpenAI safety lead Houda responds that she is less worried about existential risk today than in 2022 despite far more capable models — the biggest uncertainty then, whether labs would seriously invest in alignment, has meaningfully resolved. She argues p(doom) is endogenous, not an exogenous constant: catastrophe probability depends on what labs, governments, researchers and society actually do, so numbers without those assumptions tell us little. She still believes AI's expected benefits outweigh costs, conditional on getting it right.

Related event: Anthropic CEO Dario Amodei Calls for Slowing Down Frontier AI(194 posts)→

Original post →

More from AGI Musings

AGI Musings channel →