Anthropic pitches a federal AI bill that would let a judge halt dangerous models

Anthropic representatives have pitched US lawmakers a draft federal bill on frontier AI safety, Politico reported on 16 October 2026, citing three people [1][2]. Under the draft, the government could block a model's release or use only for a foreseeable, imminent risk of more than $1 billion in damages or of killing or seriously injuring more than 50 people [1].

What would Anthropic's draft bill allow the government to do?

The Commerce Department could ask a federal judge for the District Court for the District of Columbia to suspend a model, Reuters reported, citing Politico [1]. The draft explicitly bars any other US department, agency or officer from suspending development, deployment or internal use [1].

Independent evaluators registered with the Commerce Department would periodically assess each lab and its frontier models [1]. Labs could pay them, as long as payment is not tied to the outcome, and evaluators who find an "imminent enumerated risk" would have to tell Commerce within 72 hours [1].

The spokesperson said the bill also requires developers to test frontier models, be transparent about findings and get them independently evaluated [1].

How does the draft relate to Anthropic's own framework?

An Anthropic spokesperson told Reuters the bill represents proposals in the company's Advanced AI Framework, released earlier this year [1]. That framework covers models trained with more than 10^25 floating-point operations at companies with more than $500 million in AI revenue or more than $1 billion in AI research spending [3].

The draft would also override certain state laws, including California's, Reuters reported [1]. Anthropic wrote in June that "absent a strong federal law, states should retain full authority to legislate" [1]. Its framework says Congress should not preempt state law unless a federal law is at least as strong as the framework [3].

Charlie Bullock of the Institute for Law & AI called the proposal "pretty reasonable" but said on X he considers it "weaker" than the FRONTIER Act, a competing House bill introduced this summer [1].

Why does Anthropic want a federal rulebook now?

Washington forced Anthropic to suspend its most advanced models for nearly three weeks in June, and chief executive Dario Amodei called on the industry last month to slow down development, Reuters reported [1]. Congress has yet to pass a federal AI law and is out of session until after the November midterm elections [1].

Co-founder Jack Clark told IMF and World Bank meetings in Bangkok on 17 October 2026 that international cooperation was needed for "unusual and serious rules" for AI [4]. He cited an internal Anthropic forecast that US GDP could reach $44 trillion by 2030 under certain conditions, against just over $30 trillion now [4].

Our read

Opinion: A judge-only pause with a bar on every other agency reads as a guard against another June-style suspension as much as against catastrophe. It is still more concrete than any other frontier lab has put on paper.

What we could not verify

• We could not open Politico's report or the draft text; the bill's details here come from Reuters' account of that report.
• Anthropic has not published the draft, and it is not known which lawmakers received it.
• Clark's $44 trillion figure is a forecast from an internal Anthropic model, not an independent estimate.