The White House Credit: Daniel Schwen / CC BY-SA 3.0 via Wikimedia Commons Senate negotiators are drafting a bill that would let the US government stop the release of an AI model. Courtney Rozen reported the terms for Reuters. Her sources are two Senate aides, a person familiar with the deliberations and a lobbyist involved in the talks. A duty of care, and a power to block The bill would create a duty of care for AI developers, according to two Senate aides.
Companies would have to design their products with the goal of preventing catastrophic risks. It would also give the government the power to block the release of certain models judged unsafe. A company could challenge that decision in federal court. How much power the government would hold is still under discussion.
The bill would apply to models with the most advanced capabilities. Reuters named Google, Anthropic and OpenAI as the US companies holding those models. Who is writing it Three senators are running the talks. John Thune is Senate majority leader.
Ted Cruz chairs the Senate Commerce Committee, which oversees the Commerce Department and its in-house AI safety researchers. Amy Klobuchar is the senior Democrat in the room. Cruz wrote on X that he is working with Klobuchar and Thune on catastrophic risks involving biological or nuclear threats. He said this month that the committee may hold a markup.
He cancelled one scheduled before the August recess, saying he wanted to focus on child safety legislation. Klobuchar said in a statement that any agreement has to make developers work with government experts to verify and test models. The US cannot allow the release of dangerous models, she added, including those that evade the control of their developers. Thune’s spokesperson declined to comment to Reuters.
The fight is over who runs the tests Maria Cantwell is the ranking Democrat on Commerce and is not in the negotiating trio. Alexandra Kelley reported for Nextgov that she is blocking over the testing framework. A Democratic committee aide described the draft this way. Companies would run the safety tests themselves, then present the results to the Commerce secretary for deployment approval.
Cantwell wants the models sent to federal agencies instead, including the national laboratories and national security agencies. The aide called the draft “primarily a voluntary standard type situation”. Cantwell’s office has offered alternative language to Klobuchar’s, which the aide described as receptive. Cruz’s approval is the deciding factor.
A second person familiar with the matter told Nextgov that Cantwell, Anthropic and AI safety groups are “refusing to play ball”. Anthropic did not answer Nextgov’s request for comment. Cantwell set out her position on X. Meaningful legislation, she wrote, would send the most powerful models to scientists at the national laboratories.
They would assess whether the models could enable sophisticated cyberattacks, or aid the development of biological or nuclear weapons. The clause that would override the states Part of the bill would stop states enforcing their own laws on certain model risks, the two Senate aides told Reuters. One gave an example: a bad actor using an AI system to design nuclear or biological weapons. That matters because the states are where the rules currently sit.
California passed SB 53, the law OpenAI first opposed and then asked the state to toughen. New York passed the RAISE Act. Colorado’s own legislature gutted its law this year, after a lawsuit. Cantwell connected the preemption clause to last year’s fight.
Republicans are waking up to the need for tough AI safety rules, she wrote on X, and she welcomes the urgency. She called it a sharp reversal from the attempt to impose a ten-year moratorium on state AI regulation. The Senate stripped that moratorium from a bill by 99 votes to one in July 2025. Federal rules for reviewing frontier models already exist, and they are not public.
TNW reported on 9 September that the White House has not released its framework, and that the companies who have read it are the largest ones. The calendar is the constraint The midterms are on 3 November. The Senate is expected in Washington for three more weeks. The House is scheduled for one, returning on 9 November.
Four House Democrats asked the Speaker to change that. Sam Liccardo drafted a letter and George Whitesides, Lori Trahan and Ted Lieu co-signed it. It calls for the House to return immediately and stay in session until it advances bipartisan AI safeguards. Andrew Solender obtained it for Axios.
The letter points to bills already filed, including the AI Kill Switch Act from Lieu and Nathaniel Moran, and the FRONTIER Act from Trahan and Jay Obernolte. Mike Johnson answered on NBC’s Meet the Press on 13 September. He said he would recall the House and hold a vote if there were a solution to vote on. He wants a summit first, with the president, congressional leaders and the heads of the major labs.
He is putting a data centre energy bill on the floor this week instead, Axios reported. The White House is not asking for this Asked in Dallas last Thursday whether he had concerns about AI’s existential risks, Donald Trump said he had none. Speaking in Ireland on Sunday, he described the warnings as coming from negative forces raising things that will not happen, and said whoever wins AI wins. TNW covered those remarks on Monday.
David Sacks, his former AI czar, has told the labs to slow themselves down without legislation. TNW reported that argument on Sunday. Johnson used the same reasoning on Meet the Press. What is not known Nobody outside the talks has seen the text.
That leaves three things open: the scope of the government’s blocking power, the shape of the court challenge, and the reach of the preemption clause. Klobuchar and Thune wrote a bill together in 2023, the Artificial Intelligence Research, Innovation, and Accountability Act. It did not pass. The Democratic aide told Nextgov that it informs the current draft but that this is likely to be a fundamentally different bill.
Leave a Reply