Anthropic proposes federal AI safety bill targeting catastrophic risks

Anthropic has privately circulated draft federal legislation that would sharply limit when Washington can stop a dangerous AI model, according to a Politico report relayed by LatestLY and Business Today. Under the draft, the Commerce Department could seek to block a model only when it poses a "foreseeable and imminent danger" of more than $1 billion in damage, or of killing or severely injuring more than 50 people. The covered risks are biological weapons, cyberattacks and a model evading its developer's controls.

The most consequential provision is procedural. Commerce could not shut a model down on its own: it would have to ask a federal judge in the U.S. District Court for the District of Columbia, and the draft bars other agencies from halting a model. Independent evaluators registered with Commerce would have to report an imminent enumerated risk within 72 hours. The draft also adds transparency reports, whistleblower protections and critical incident reporting, and it would preempt some state rules, with language that mirrors California's SB 53.

For the AI industry this is a bid to set the terms of federal preemption before Congress writes its own. A single national standard with a high, court-supervised bar for intervention would reduce the risk of a patchwork of state laws and of sudden shutdown orders, which matters for developers and for the cloud and chip companies whose revenue depends on frontier model deployment. Critics are likely to argue that the thresholds are high and that the proposal comes from a company it would regulate.

What to watch: whether any lawmaker sponsors the text (it has not been introduced in either chamber and no recipients were named), how it interacts with Senate and House AI bills under discussion, and whether rival labs back or oppose a court-gated standard. The timing also lands as Anthropic prepares for its planned public listing.