Anthropic has been circulating a draft federal bill on AI safety that it hopes Congress will take up, according to a Politico report cited by TechCrunch. The proposal sets out narrow scenarios in which an AI model's release could be stopped before distribution.
The trigger is a "foreseeable, imminent risk" that more than 50 people could die, or that property damage could exceed a billion dollars, stemming from a biological weapon, a model escaping its developer's controls, or a cyberattack. A company spokesperson confirmed the draft to Politico, saying it reflects the Advanced AI Framework Anthropic has been promoting with federal and state lawmakers since June.
Anthropic AI safety bill: how the trigger would work
Under the draft, catastrophic-risk analysis would be handled by "independent evaluators" registered with the US Commerce Department. Those evaluators could be paid by the AI labs themselves, as long as the payment is not contingent on the evaluation's results, Politico reported. How that arrangement would keep payments from influencing a verdict is not spelled out.
If evaluators find an imminent catastrophic risk, they would have 72 hours to notify the Commerce Department. The department would then have to ask a D.C. judge to halt the model's release. The bill would expressly bar any federal agency from halting distribution on its own discretion.
- Risk threshold: 50+ potential deaths or $1B+ in property damage
- Named scenarios: biological weapon, model evading developer controls, cyberattack
- Evaluators: independent, registered with the US Commerce Department, may be paid by AI labs if payment is not tied to results
