 |
ILLUSTRATIVE IMAGE: A modern data center, representing AI safety proposals and technology regulation.
Anthropic Proposes AI Safety Bill With Limits on Government Intervention
WASHINGTON — October 16, 2026
Artificial intelligence company Anthropic has presented US lawmakers with a draft bill that would establish rules for the safety and security of advanced AI systems, Politico reported on Friday, citing three people familiar with the discussions. The proposal would set specific conditions under which the US government could seek to stop the release or use of a powerful AI model.
Under the draft legislation, government intervention would be permitted only if a model posed a foreseeable and imminent risk of causing more than $1 billion in damage, or killing or seriously injuring more than 50 people. The reported thresholds would apply to risks arising from the creation or release of a biological weapon, an AI model escaping its developer's controls, or a cyberattack.
Anthropic told Reuters that the proposed bill reflects ideas in its Advanced AI Framework, which the company released earlier in 2026. A company spokesperson said the proposal would give the government a way to prevent dangerous deployments of the most powerful AI models while requiring developers to test their systems, disclose findings and obtain independent assessments.
The draft would also establish a system of independent evaluators registered with the US Department of Commerce. These evaluators would periodically assess AI developers and their most advanced models, according to Politico's report.
AI companies could pay for these assessments, provided the payments were not linked to whether the evaluators reached a favorable or unfavorable conclusion. If an evaluator identified an imminent risk covered by the proposed law, the evaluator would have to notify the Commerce Department within 72 hours of completing the assessment.
The proposal would place limits on which parts of the government could seek to halt an AI model. According to Politico, the Commerce Department could ask a judge to suspend a model after determining that it posed the specified risk, while other federal departments, agencies and officials would be barred from independently suspending its development, deployment or internal use under the draft's terms.
A federal judge in the US District Court for the District of Columbia could order a suspension, the report said. The Department of Commerce had not immediately responded to a request for comment.
The proposal comes amid wider debate over how governments should oversee increasingly capable AI systems without unnecessarily restricting their development. Anthropic's draft seeks to define the circumstances that would justify intervention, while setting testing and independent evaluation requirements for developers of advanced models.
The reported bill remains a proposal, and the supplied report does not establish that lawmakers have introduced or approved it as legislation.
