Anthropic wrote its own AI safety bill for Congress

Per POLITICO, which obtained a confidential draft, the government could block a model's release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people. Three people familiar with the effort say Anthropic reps have quietly pitched the bill on Capitol Hill.

The risk has to come from one of three places: a biological weapon, a model evading its developer's controls, or a cyberattack. The language mirrors California's AI laws, including SB 53, signed last year.

An Anthropic spokesperson said the bill tracks the company's Advanced AI Framework, which it has pushed with federal and state lawmakers since June. The report doesn't say which lawmakers saw the draft or whether anyone plans to introduce it.