Anthropic Proposes AI Safety Legislation to Prevent Model Releases with Catastrophic Risks
Anthropic has proposed an AI safety bill that could block the release of models if they pose a foreseeable risk leading to over 50 deaths or $1 billion in property damage. The draft legislation outlines scenarios involving biological weapons, model control evasion, or cyberattacks as catastrophic risks and suggests using independent evaluators registered with the U.S. Commerce Department for assessments. The bill would require these evaluators—potentially funded by AI labs—to notify the Commerce Department within 72 hours if a risk is identified, prompting judicial review to halt releases. The proposal aligns with California's Senate Bill 53 and diverges from voluntary oversight models advocated by other industry players like Nvidia’s CEO Jensen Huang. While Anthropic emphasizes its framework as part of ongoing advocacy efforts since June, the bill raises questions about potential conflicts of interest in evaluator funding arrangements.
Topics: AI safety artificial intelligence government regulation policy proposal