Anthropic has quietly started lobbying for a new AI safety law. The proposal would let the U.S. government stop an AI model from being released, but only if the model could cause either 50 or more deaths or at least $1 billion in damages. The bill would put the Department of Commerce in charge of the decision and would use an independent evaluator to check the lab’s risk assessments.
How the bill would work
According to Politico, a third‑party evaluator registered with the Commerce Department would periodically review Anthropic’s labs and its AI models. The lab could pay the evaluator, but the payment could not influence the evaluator’s independence. Anthropic says the approach follows its “Advanced AI Framework,” which it has been promoting to lawmakers since June.
Why it matters
If passed, the law would give the Commerce Department the power to block risky AI models, but only after a very high threshold of harm is demonstrated. For everyday users, this means most AI tools would stay available unless a clear, large‑scale danger emerges, leaving safety concerns largely to be addressed by the companies themselves.
