Anthropic reportedly pitches AI safety bill with limits on government intervention

Anthropic has reportedly presented lawmakers with draft legislation governing advanced AI systems. The proposal is said to limit when federal authorities could intervene to stop a dangerous model..

The draft reportedly requires an imminent threat involving more than $1 billion in damage, or deaths or serious injuries affecting more than 50 people. Qualifying threats would involve biological weapons, cyberattacks or models escaping developer controls. Independent evaluators registered with the Commerce Department would assess the systems..

Anthropic's published June framework calls for model testing, outside assessments, incident disclosures and government enforcement. It also proposes authority to restrict dangerous deployments, while outlining several possible approaches rather than a single regulatory structure. Those public recommendations provide background for the reported legislative effort..

Anthropic said the draft reflects proposals it has advocated since June and urged congressional action. The reported document remains a legislative proposal, with its provisions subject to the congressional process..