United States: One week before mid-term elections, pressure rises for a federal AI law, without consensus in Congress

Two thirds of voters are calling for more regulation and AI agents have already accessed bank accounts. But negotiations in the Senate are blocked and the House will not vote before the elections.

In the run-up to the November 3 election, AI regulation is gaining ground in the American debate. A Navigator Research poll conducted in early October indicates that 66% of registered voters want more supervision, six points more than in December 2025; Reuters/Ipsos measures support for stricter federal regulation among 80% of Democrats and 61% of Republicans. However, AI remains secondary in priorities: only 21% cite it among their top five topics, far behind inflation.

Concrete incidents, a divided Congress

Major U.S. banks have begun restricting third-party AI agents after cases of agents accessing customer accounts beyond instructions received, prompting a reaction from Senator Mark Warner. On the legislative side, bipartisan discussions in the Senate have stalled since late September, particularly on the issue of mandatory pre-market testing. In the House, a bill by Reps. Jay Obernolte and Lori Trahan would allow the Commerce Department to suspend the development or deployment of an advanced model with an "imminent catastrophic risk," but the chairman of the relevant committee refuses to commit to a vote this year.

The preemption of state laws, such as those in California and New York, remains one of the main sticking points. Observers believe that a Democratic victory could accelerate legislation in 2027, or at least increase hearings and requests for documents addressed to developers.