Anthropic pitches its own AI safety bill to US lawmakers

Anthropic representatives quietly pitched Capitol Hill lawmakers on a draft bill to regulate frontier artificial intelligence safety and security.

The proposed legislation would allow the government to block a model's release only if it poses an imminent risk of over $1 billion in damages or significant casualties.

Independent registered evaluators would assess labs and report risks to the Commerce Department within 72 hours of completing their assessments.