Anthropic has circulated a draft bill that asks Congress to let the U.S. Commerce Department stop the distribution of an AI model when an "imminent catastrophic risk" is identified. The bill defines that risk as one that could kill more than 50 people or cause over a billion dollars in property damage. The danger must come from a biological weapon, the model escaping its controls, or a cyber‑attack.
How the bill would work
The proposal calls for "independent evaluators" registered with the Commerce Department to assess a model’s risk. Evaluators could be paid by AI labs, but the payment could not be tied to the outcome of the assessment. If an evaluator flags a risk, they have 72 hours to notify the Commerce Department, which must then ask a D.C. judge to order a halt on the model’s release. The bill bars any federal agency from stopping a model on its own.
Industry response
Anthropic says the bill reflects its Advanced AI Framework, which it has been promoting to lawmakers since June. The company has not commented further. Other AI firms have pushed for voluntary internal monitors rather than mandatory oversight, and Nvidia’s CEO has argued the industry needs no new laws.
Why it matters
If adopted, the bill could stop the release of AI systems that pose severe safety threats, giving the public a legal backstop against extreme harms. For developers, it would add a new compliance step before launching powerful models. It remains unclear how independent the evaluations would be or how quickly courts could act in practice.
