Anthropic pitches Congress its own frontier-AI bill: models could only be halted above 50 deaths or $1B in damage
What's happening? Anthropic has been quietly pitching US lawmakers on a draft federal bill to regulate frontier AI safety and security, Politico reported on 16 October 2026, citing three people familiar with the effort and a copy of the confidential draft. Anthropic confirmed the draft is circulating, saying it "represents the proposals outlined in Anthropic's Advanced AI Framework, which we have been advocating for with federal and state lawmakers since June," and called them "sensible policies we believe Congress should act on."
What the draft says (as reported by Politico): 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, and only when that risk stems from one of three sources: the creation or release of a biological weapon, a model evading its developer's controls, or a cyberattack. Independent evaluators registered with the Department of Commerce would periodically assess labs and their frontier models; labs could pay them as long as payment does not depend on the outcome. Evaluators finding an "imminent enumerated risk" would have 72 hours to report it to Commerce, which could then ask a federal judge in Washington, DC to suspend the model. The text bars any other US department, agency or officer from suspending a frontier model under the Act. Labs would also have to keep safety frameworks, publish transparency reports, protect whistleblowers and report critical incidents, and the bill would preempt some state rules on risk disclosure, third-party evaluation and incident reporting.
The bigger fight: Congress has still not passed a federal AI law and is out of session until after the November midterms. Anthropic's draft joins a crowded field: the House FRONTIER Act from Reps. Lori Trahan and Jay Obernolte, a catastrophic-risk bill being negotiated by Senate Majority Leader John Thune with Sens. Amy Klobuchar and Ted Cruz, and calls from Rep. Greg Casar and Sen. Bernie Sanders to pause frontier development altogether. The preemption clause is the flashpoint: states such as California (SB 53, which Anthropic supported and whose language the draft tracks in places) and New York have been the ones actually writing frontier-model rules, and Anthropic itself wrote in June that, absent a strong federal law, states should retain full authority to legislate. The company also has first-hand experience of discretionary federal action: AFP notes Washington forced it to suspend its most advanced models for nearly three weeks in June under export controls. Speaking at the IMF/World Bank meetings in Bangkok the same day, co-founder Jack Clark said it was remarkable that the toys he buys for his children are more regulated than AI.
Reaction and interpretation: Online, critics fixated on the 50-person threshold and called the draft self-serving, since the companies it would regulate wrote it. Some legal readers argue the opposite effect: a congressionally approved route to a court-ordered block could make a federal halt more likely than under today's law, and the preemption section as drafted appears not to cover state suspension of models. Politico notes that none of the AI-agent incidents labs disclosed this summer appear to have come close to the bill's thresholds. These are interpretations of a leaked draft; the final text, sponsors and timing are not public.
Why it matters: A frontier lab heading for a public listing is now offering lawmakers a ready-made answer to the question every AI bill must settle: who can stop a model, how bad the risk must be, and whether federal rules override the states. Whatever passes will set the liability floor for the whole industry, not just Anthropic.
Sources: Politico (16 Oct 2026); Reuters via Investing.com; AFP via Dawn; Mashable; LessWrong analysis of the draft.
