Anthropic Privately Lobbies U.S. Lawmakers With Self-Drafted AI Safety Legislation

Anthropic has drafted a self-developed frontier AI safety regulatory bill and privately promoted it to U.S. members of Congress. The draft sets a risk threshold for government intervention in model releases: when an AI model presents a foreseeable immediate risk that could cause over $1 billion in property damage, or result in more than 50 deaths or severe injuries, with the risk stemming from the model's ability to create and disseminate biological weapons, break free of developer control, or launch cyberattacks, the government may block its release or use. The draft requires assessment bodies to report to the Department of Commerce within 72 hours after identifying a "specified immediate risk"; after confirmation by the Department of Commerce, it may apply to the court for an injunction, while prohibiting other government agencies from unilaterally suspending the research, development, deployment, or internal use of models. The draft also mandates developers to test models and disclose results, and undergo regular reviews by independent assessment bodies.

Source: PANewsRead original