Anthropic Proposes Federal AI Safety Bill to Congress

Anthropic has formally presented a draft AI safety bill to Congress that establishes a risk-based framework for federal intervention on AI model deployment. The proposal, pitched to lawmakers, would empower the government to halt the release of an AI model if it poses a foreseeable, imminent catastrophic risk.

Anthropic representatives discussed the draft AI safety bill with US lawmakers. The company confirmed the draft bill via a spokesperson. "This bill represents the proposals outlined in Anthropic’s Advanced AI Framework, which we have been advocating for with federal and state lawmakers since June," an Anthropic spokesperson said.

Risk-Based Deployment Triggers

The bill defines specific catastrophic risk thresholds requiring federal action, including threats to over 50 lives or $1 billion in damages from biological weapons, loss of control, or cyberattacks. A model's release could be blocked if there is a foreseeable, imminent risk of it killing or seriously injuring more than 50 people or causing over one billion dollars in property damage. The threat must result from the model creating or releasing a biological weapon, operating beyond human control, or enabling a major cyberattack.

Independent Evaluation Requirements

The legislation mandates AI companies to allow routine independent evaluations of frontier models by Commerce Department-registered assessors. Developers would be required to test frontier AI models, be transparent about their findings, and submit to these independent reviews. Evaluators would register with the U.S. Commerce Department and periodically evaluate AI labs and their frontier models. Labs could pay the evaluators, provided the payment was not directly tied to the outcome of the evaluation.

Emergency Notification and Enforcement

Evaluators must notify the Commerce Department within 72 hours of identifying imminent catastrophic risks, with the department seeking judicial intervention to halt deployments. Upon completing an assessment, evaluators have 72 hours to report an "imminent enumerated risk" to the Commerce Department. The Commerce Department would then request a judge from the District Court for the District of Columbia to order the model's suspension. The bill explicitly prohibits any other US government department, agency, or officer from unilaterally suspending a model's development, deployment, or internal use.

The bill would also mandate that AI makers publicly share the findings from their model testing and the outcomes of independent evaluations.