Anthropic Pitches Draft AI Safety Bill to Congress, Raising Intervention Thresholds and Preempting State Laws

According to an exclusive Politico report, representatives from AI giant Anthropic have quietly pitched Capitol Hill lawmakers on a draft bill to regulate frontier AI safety and security. The confidential document marks a notable effort by the frontier lab to shape the rules governing its own technology as Congress weighs how to respond to the fast-moving industry.

Key provisions of the draft include limiting government power to block a model's release or use only when there is a "foreseeable, imminent" risk of more than $1 billion in damages or death/serious injury to more than 50 people. This risk must stem from biological weapons, model evasion of controls, or cyberattacks. While mirroring language found in California's SB 53, the bill includes strong clauses prohibiting other US government departments from suspending development or deployment, centralizing such authority with federal courts and the Commerce Department.

The proposal also requires independent evaluators to register with the Commerce Department and periodically assess labs and their models. If an imminent risk is found, evaluators must notify Commerce within 72 hours. Commerce could then ask a judge to halt a model.

The pitch has drawn sharp criticism from AI policy experts. Zvi Mowshowitz noted that compared to the previously proposed FRONTIER Act, Anthropic's draft features significantly weaker enforcement provisions, particularly regarding emergency shutdown authorities and rulemaking. Critics argue Anthropic appears more concerned about protecting its business interests from government overreach than ensuring robust public safety mechanisms, attempting to leverage the current political climate to preempt important state laws.