Anthropic pitches AI safety bill setting thresholds for model suspensions

Anthropic has approached lawmakers on Capitol Hill with proposed legislation on frontier AI safety and security, according to an October 16 media report citing three people familiar with the effort. The draft would limit when the government could prevent a model from being released or used.

Government intervention would require a foreseeable and imminent threat involving damages above $1 billion, or death or serious injury to more than 50 people, the report said. The threat would need to arise from a biological weapon being created or released, an AI model escaping developer controls, or a cyberattack.

The proposal calls for independent evaluators to register with the Commerce Department and periodically assess AI laboratories and their frontier models. Laboratories would be allowed to pay for those assessments, but compensation could not be linked to the findings, according to the report.

Evaluators identifying an imminent threat covered by the legislation would have to inform Commerce within 72 hours after completing an assessment. If the department determined that a risk existed, it could ask a judge to halt the model.

A judge on the District Court for the District of Columbia could order a model suspended, the report said. The draft would explicitly prevent other U.S. government departments, agencies or officers from suspending a model's development, deployment or internal use.

Neither Anthropic nor the Commerce Department immediately responded to requests for comment.