Anthropic quietly hands Congress its own draft AI safety bill, with a high bar before the government could block a model

An AI company has written the law it wants to live under, and it has begun handing the draft to Congress. Anthropic representatives have quietly pitched Capitol Hill lawmakers on a draft bill to regulate the safety and security of frontier AI models, Politico reported on Friday, citing three people with knowledge of the effort. Reuters reported the pitch on October 16, and an Anthropic spokesperson confirmed to Reuters that the draft puts into legal language the proposals in the company's Advanced AI Framework, released earlier this year.

The core of the draft is a threshold. The government could block a model's release or use only if the model poses a foreseeable and imminent risk of causing more than 1 billion dollars in damages, or of killing or seriously injuring more than 50 people. The risk would have to come from one of three sources: the creation or release of a biological weapon, a model evading its developer's controls, or a cyberattack. Below that line, the federal government would have no power to stop a deployment under the bill.

Around that threshold the draft builds a testing regime. Developers of frontier models would have to test them, publish what they find, and submit to evaluation by independent evaluators registered with the Department of Commerce. Labs could pay the evaluators, provided the payment is not tied to the outcome. If an evaluator finds an imminent risk of the kind the bill names, the evaluator must report it to Commerce within 72 hours of finishing the assessment. Commerce could then ask a federal judge in the District of Columbia to suspend the model. The draft expressly bars every other department, agency or officer of the government from suspending development, deployment or internal use on its own authority.

The Department of Commerce did not immediately respond to a Reuters request for comment, and no lawmaker has yet introduced the draft as a bill. What exists is a company's preferred shape for the first federal AI safety law, circulated quietly while Washington still has none.