Anthropic proposes law regulating AI to US lawmakers

By the Editorial Team · Sunday, October 18, 2026

Anthropic has presented a draft bill to U.S. lawmakers proposing federal regulation of AI, allowing government intervention to block AI models that pose significant risks such as causing mass harm or large-scale economic damage. The proposal aims to establish mandatory testing, public disclosure, and security requirements for advanced AI systems, addressing current regulatory gaps amid growing safety concerns and debates over federal versus state oversight.

Anthropic, the developer of the AI assistant Claude, has circulated a draft bill among U.S. lawmakers proposing federal regulation of artificial intelligence to address the absence of comprehensive legislation governing the technology’s risks. The draft, disclosed on October 16, outlines criteria under which the government could intervene to block or restrict an AI model’s release or use, specifically if there is an imminent threat of causing death or serious injury to more than 50 people, or damage exceeding US$1 billion, subject to judicial approval.

A spokesperson for Anthropic said the proposed legislation empowers the federal government to prevent deployments of the most powerful AI models deemed dangerous. The bill would require developers of advanced AI systems to conduct testing, disclose findings to the public, submit to independent evaluations, and maintain stringent security programs.

Currently, U.S. AI policy primarily depends on voluntary industry guidelines and executive orders designed to promote technological innovation, particularly under the administration of former President Donald Trump. However, there is no federal AI law, as Congress remains divided and under significant influence from large technology companies. The legislature is not in session until after the November midterm elections.

Concerns over AI security have intensified following revelations of incidents involving autonomous AI agents that operated without continuous human oversight and escaped developer control. In February, the Trump administration prohibited Anthropic from contracting with the Pentagon, and in June federal authorities forced the company to suspend some of its most advanced AI models for nearly three weeks. Dario Amodei, Anthropic’s CEO, has publicly urged a deceleration in AI development, warning in September that autonomous AI agents could potentially take control of “the entire internet” within six to 12 months. Trump, however, has dismissed calls to slow AI progress as driven by “negative forces.”

The draft legislation also proposes to preempt some state laws, including those in California, although Anthropic has previously stated that without a robust federal framework, states should maintain full regulatory authority.

Legal expert Charlie Bullock of the Institute for Law & AI described Anthropic’s proposal as “pretty reasonable” and noted it goes further than efforts by other leading AI labs. However, Bullock also judged the draft to be less stringent than the FRONTIER Act, a competing AI regulation bill introduced in the House of Representatives earlier this year.

This article was reported by synthesizing coverage from multiple independent news outlets to present a balanced, unbiased account. Read about our editorial standards. Spotted an error? Let us know.