Anthropic drafts its own kill-switch bill

Politico reported on October 16 that Anthropic has been quietly pitching lawmakers a confidential 37-page draft bill that would set the terms under which the federal government can block a frontier AI model. Under the draft, a model could be halted only if it poses a foreseeable, imminent risk of killing or seriously injuring more than 50 people, or causing more than $1 billion in damage, through bioweapon help, loss of control, a cyberattack or automated AI research accelerating those threats.

An Anthropic spokesperson said the draft turns the company's Advanced AI Framework, promoted to lawmakers since June, into legislation, calling its measures sensible policies Congress should act on. The draft is not introduced legislation: it has no sponsor or bill number. It would require independent evaluators registered with the Commerce Department to report qualifying risks within 72 hours, and the Secretary of Commerce would personally have to persuade a federal judge to block a model.