Anthropic privately offers lawmakers a draft bill on frontier AI safety

what happened

Anthropic has quietly approached US lawmakers with a draft bill on safety and security rules for frontier AI, Politico reported on Friday, citing three people familiar with the talks. Under the proposal, the government could stop a model's release or use only when a risk is foreseeable, imminent and tied to damages above $1 billion or to deaths or serious injuries of more than 50 people. Qualifying risks would have to involve a biological weapon, a model slipping out of its developer's control, or a cyberattack.

Independent evaluators registered with the Commerce Department would review labs and their frontier models on a regular basis. Labs could pay them, as long as the fee does not depend on the result. If evaluators find an imminent risk, they must notify Commerce within 72 hours. Only Commerce could ask a court to halt a model, and other federal bodies would be barred from suspending it. A federal judge in the District of Columbia could order such a suspension. Anthropic and the Department of Commerce did not immediately respond to requests for comment.

why it matters

For markets this is an indirect regulatory story: rules for frontier AI models could affect how quickly large AI developers release new systems, which matters for AI-linked shares and cloud spending. The proposal is still a private draft, so its path through Congress is unclear.

tags

instruments