Anthropic has circulated a draft bill proposing a formal review process for advanced AI models that could pose an imminent, catastrophic risk. The proposal would allow a model's release to be paused if an assessment finds a foreseeable threat involving more than 50 deaths or over $1 billion in property damage.
The draft identifies three possible sources of such risk: the creation or release of a biological weapon, an AI model evading its developer's controls, or a cyberattack. Its thresholds resemble those in California's Senate Bill 53, which also defines catastrophic risk using measures of potential harm.
Independent assessments and court oversight
Under the proposal, evaluators registered with the U.S. Department of Commerce would assess a model's potential risks. AI companies could fund the evaluations, provided payment was not tied to the findings. If evaluators identify an imminent catastrophic risk, they would have 72 hours to notify the department.
The department would then need to ask a Washington, D.C., judge to halt the model's release. The draft would not give federal agencies authority to stop distribution on their own.
Anthropic says the draft reflects measures in its Advanced AI Framework, which it has been advocating to federal and state lawmakers. The proposal would put legal review and independent assessment at the center of decisions about releasing high-capability AI.
If adopted, the framework could help shape how public oversight and technical evaluation work together as AI systems become more capable.
