Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports

The proposed bill outlines specific triggers for government intervention, focusing on extreme risks. A model could be halted if it poses a foreseeable, imminent risk of over $1 billion in damages or could cause more than 50 deaths or serious injuries. These risks would need to stem from the creation or release of a biological weapon, the model evading its developer’s controls, or a cyberattack, according to Politico.

The framework also includes requirements for developers to test frontier AI models, be transparent about their findings, and undergo independent evaluations. These independent evaluators, who would register with the Commerce Department, would periodically assess labs and their models. They would report any "imminent enumerated risk" to Commerce within 72 hours of their assessment.

A key element of the draft is the limited scope of intervention. While the Commerce Department could request a federal judge to halt a model, the bill explicitly prohibits any other US government department, agency, or officer from suspending development, deployment, or internal use, according to the report. This suggests an attempt to centralize and narrowly define the authority for AI regulation, potentially limiting broader governmental oversight.