Anthropic proposes AI safety bill to US lawmakers

Illustrative picture of artificial intelligence. Credit: Havard Business School

Anthropic, the creator of the AI assistant Claude, confirmed on Friday it is circulating a draft bill among US lawmakers aimed at regulating artificial intelligence to address the lack of federal legislation on the risks posed by the new technology.

According to the draft, first reported by Politico on Friday, Washington could block a model’s release or use if there is an imminent risk of killing or seriously injuring more than 50 people, or causing more than $1 billion in damages, but only with a judge’s approval.

A company spokesperson told AFP that the proposal “grants the federal government the ability to block or deter dangerous deployments of the most powerful models.”

“It requires frontier developers to test the most powerful models, be transparent to the public about their findings, submit them to independent evaluation, and maintain a robust security program,” the spokesperson said.

AI policy in the United States under President Donald Trump has relied primarily on voluntary industry pledges and executive orders intended to sustain an AI boom that tech leaders say is driving the country’s economy.

Congress, politically divided and heavily lobbied by Big Tech, has yet to pass a federal AI law, and is out of session until after next month’s midterm elections.

Security became a hot topic after recent revelations of security incidents caused by AI agents — autonomous software that operates without constant human supervision — that escaped the control of their developers.

Anthropic, which was banned from the Pentagon by the Trump administration in February, was forced by Washington in June to suspend its most advanced models for nearly three weeks.

• US Supreme Court to hear Nigerian ‘gunmen’s cook’ asylum case
• AI can’t replace human wisdom, says expert
• FG deploys AI, digital audits to monitor immigration officers

Anthropic CEO Dario Amodei called on the industry last month to slow down development, fearing that a group of AI agents could take control of “the entire internet” within six to 12 months.

Trump has dismissed the numerous calls to slow down the technology as “negative forces.”

The proposed legislation would also override certain state laws, such as those in force in California.

Anthropic wrote in June that “absent a strong federal law, states should retain full authority to legislate.”

Charlie Bullock of the Institute for Law & AI think tank praised the proposal as “pretty reasonable,” and “more than other frontier labs have done.”

But he said on X he considers it “weaker” than the FRONTIER Act, a competing bill introduced in the House of Representatives this summer.