Anthropic secretly lobbies Congress: Models must face over $1 billion in losses or risk of 50 casualties before the government can intervene

Beating AI News Flash: According to Politico, Anthropic is privately lobbying U.S. lawmakers on a draft frontier AI safety bill that would set strict thresholds for government intervention in AI models. Under the draft, the government could seek to halt a model only when it faces a foreseeable and imminent specific risk that could cause more than $1 billion in losses or lead to the death or serious injury of more than 50 people.

The draft limits regulatory triggers to three categories of risk: AI models being used for the development or proliferation of biological weapons, models escaping developer control, and models enabling cyberattacks. The bill would also require AI developers to conduct transparent testing and undergo independent third-party evaluations. If an evaluating institution identifies relevant risks, it must report to the U.S. Department of Commerce within 72 hours, and the Commerce Department may then apply to a federal court to suspend the model's operations, with the court making the final ruling.

In addition, the draft would prohibit other U.S. government agencies and officials from independently halting the development, deployment, or internal use of AI models. Anthropic said the proposal stems from its advanced AI framework released earlier this year, which aims to require developers to undergo safety evaluations while giving the government the ability to block the deployment of dangerous models.

The draft is still in the lobbying stage and has not become law. If the relevant provisions are ultimately adopted, they could significantly limit the U.S. government's authority to take emergency intervention against frontier AI models.