Leaked AI safety bill from Anthropic only requires government intervention if 50+ injuries or $1 billion are at risk
Politico reported on a confidential draft AI safety bill from Anthropic that was pitched to US lawmakers. The proposal would restrict government intervention to cases where an AI model poses a foreseeable, imminent risk of either more than $1 billion in damages or serious injury or death to more than 50 people.
Under the bill’s framework, regulatory action would only be triggered if the risk stems from one of three specific scenarios: creation or release of a biological weapon, the model evading its developer’s controls, or cyberattacks. The high threshold for intervention—requiring either substantial financial damages or dozens of casualties—reflects Anthropic’s approach to balancing innovation with safety oversight.
The leaked bill represents the company’s formal proposal to lawmakers for how AI model regulation should be structured in the United States. The reporting was covered by both Tom’s Hardware and PC Gamer.