By Hamza Faisal
Anthropic is not just lobbying about AI regulation anymore — it has written the regulation itself. Representatives for the Claude maker have been quietly circulating a draft federal bill among members of the US Congress that would create the country's first nationwide regime for frontier AI safety, according to a Politico report on Friday citing three people familiar with the effort. As of the report, the draft had no identified congressional sponsor and had not been formally introduced in either chamber.
The move flips the usual script: instead of lawmakers imposing rules on a lab, a lab is handing lawmakers the rulebook — one it says reflects proposals it has been pushing since its Advanced AI Framework debuted in June.
The $1 billion bar
The draft sets an extraordinarily high threshold for government intervention. Under its terms, the federal government could seek to block a model's release or use only if the model poses a foreseeable, imminent risk of causing more than $1 billion in damages — or of killing or seriously injuring more than 50 people. Anything short of that, and the model ships.
Even then, the risk has to come from one of three enumerated categories: the creation or release of a biological weapon, the model escaping the control of the company that built it, or a cyberattack. Consumer-product harms, misinformation, and labor-market disruption would fall entirely outside the bill's reach.
Commerce gets the keys — everyone else gets locked out
Enforcement would flow through a single channel: the Commerce Department. Independent evaluators would register with Commerce and periodically assess frontier AI companies and their most capable models; labs could pay those evaluators, provided the payment is not tied to the evaluation's outcome. If an evaluator finds an imminent enumerated risk, it must notify Commerce within 72 hours of completing its assessment. Commerce could then ask a federal judge in the District of Columbia to suspend the model.
Notably, the draft explicitly prohibits any other federal department, agency, or officer from independently suspending a model's development, deployment, or internal use — concentrating the off-switch in one agency's hands. An Anthropic spokesperson said the proposal grants the government "the ability to block or deter dangerous deployments of the most powerful models," while requiring frontier developers to test the most powerful models, disclose their findings, submit to independent evaluation, and maintain a robust security program.
A bill written by the company it would regulate
Early reactions are split. Charlie Bullock of the Institute for Law & AI called the proposal "pretty reasonable" and "more than other frontier labs have done," while judging it weaker than the FRONTIER Act introduced in the House this summer. The draft also appears designed to preempt a patchwork of state laws — notably California's stalled SB 1047 effort — with a single national standard.
Timing matters: CEO Dario Amodei has been warning for months that AI agents could overwhelm the internet within six to twelve months, and has called on the industry to slow down development. The bill is Anthropic's answer to that alarm — but written on its own terms, with an intervention bar set so high that almost no model short of a catastrophe would ever hit it.
The irony won't be lost on lawmakers: the company most publicly worried about runaway AI is asking Washington to regulate it — just not too much. — Hamza Faisal
