Anthropic Proposes Model Safety Legislation to United States Lawmakers

Anthropic has drafted a federal bill allowing US authorities to block risky AI models with a judge's approval.

Artificial intelligence developer Anthropic has shared a legislative draft with US politicians to establish federal oversight for advanced technology models. The proposal empowers the federal government to block the release of dangerous systems when extreme physical harm or massive financial loss is anticipated. Lawmakers have lacked comprehensive federal rules to address these emerging technological risks.

• The thresholdThe proposed rules would apply only when models present an imminent threat of severe casualties or over $1 billion in damages.
• Judicial oversightThe government would require a judge's approval before halting the release or deployment of any artificial intelligence model.
• Legislative gapThe initiative responds to the absence of federal statutes governing safety standards across the technology sector.