Anthropic's Jan Leike backs California AI report: may need to slow AI 'in a few months'

janleike · x · 2026-11-21

Jan Leike, alignment lead at Anthropic, praised a new AI policy report from the California government, highlighting mechanisms that let the government modulate the speed of development — especially for recursive self-improvement — as particularly valuable.

He quibbles with details, e.g. whether "compute devoted to automated AI R&D" is the right metric, but argues there's little time to quibble: "we might already want to actively modulate the speed of AI development in a few months." A notable endorsement of state-level slowdown mechanisms from a frontier-lab safety lead.

Original post →

More from Companies & People

Companies & People channel →