Senate AI Bill Would Impose Safety Duties, Federal Oversight

U.S. Senate negotiators are working on legislation that would establish formal safety obligations for AI developers and grant the federal government authority to block potentially dangerous AI models, according to Reuters reporting based on Senate aides and lobbyists involved in the discussions.

The proposed bill would create a "duty of care" for companies building advanced AI systems, requiring them to design products that prevent catastrophic risks. The measure targets AI models with the most advanced capabilities—systems currently developed by companies including Alphabet's Google, Anthropic, and OpenAI.

Government blocking authority with judicial review

Under the framework being discussed, the U.S. government would gain power to prevent the release of AI models deemed unsafe. Companies would retain the right to challenge such decisions in federal court, though negotiators are still determining the precise scope of government authority.

The legislation would also preempt certain state-level AI regulations. According to Senate aides, the federal law would supersede state enforcement of rules governing specific model risks, such as the potential for AI systems to enable the design of nuclear or biological weapons.

Why it matters

This represents the most concrete federal effort yet to regulate frontier AI development in the United States. By establishing baseline safety requirements and federal preemption, the bill would create a unified national framework at a time when states like California have moved to enact their own AI safety laws. The inclusion of judicial review mechanisms attempts to balance government oversight with industry concerns about regulatory overreach—a compromise that could prove critical to securing enough votes for passage.

Tight legislative calendar

The negotiations involve Senate Majority Leader John Thune, Senate Commerce Committee Chairman Ted Cruz, and Senator Amy Klobuchar, the lead Democrat in the talks. Senator Maria Cantwell, the top Democrat on the Commerce Committee, has also participated.

Timing poses a significant challenge. The House of Representatives has only one scheduled session week before the November 3 midterm elections, while the Senate has three weeks remaining. Even if negotiators reach agreement, the compressed calendar leaves little room for floor debate and passage.

In a statement to Reuters, Klobuchar said she is "continuing to work toward a bipartisan agreement on legislation for government oversight of the greatest risks posed by AI models," including requiring developers to work with government experts to verify and test models for safety.

Cantwell emphasized the need for "the most powerful AI models undergo testing by scientists and experts at our national laboratories to assess whether they could enable sophisticated cyberattacks or aid the development of biological or nuclear weapons."

The legislative push comes amid growing concerns about AI safety, including recent reports of AI agents deviating from instructions and hacking external systems, as well as public warnings from researchers who have left major AI companies citing safety concerns.

These details were first reported by Reuters correspondent Courtney Rozen.