US federal AI law: a pre-emption draft and a superintelligence ban, but nothing enacted

A federal AI pre-emption draft has no bill number yet, and lawmakers have floated a ban on superintelligence. The US still has no comprehensive federal AI law.

US federal AI law: a pre-emption draft and a superintelligence ban, but nothing enacted

One of the most discussed federal AI bills in the US has no bill number. The draft known as the Great American AI Act would set duties for frontier AI developers and override some state AI laws. As of November 2026, it has not been introduced in Congress and has had no hearing or markup, according to a guide published on 11 November by NHIMG. None of it is law.

That leaves the US where it has been for some time. A regulation overview published by iTechGuides says no comprehensive federal AI statute had been enacted as of October 2026. A separate November roundup on the Mean CEO blog says there is still no single federal AI law. Meanwhile, the state rules keep piling up.

Advertisement

A draft that calls itself an opening bid

The NHIMG guide is based on the sponsors’ own section-by-section summary. It checks each point against independent legal analysis. The draft has two main parts. One sets duties for developers of frontier AI models. The other is a pre-emption clause that would limit how far state AI laws can go.

Trahan, one of the sponsors, has described the draft as an “opening bid”, according to the guide, which cites Tech Policy Press. The guide also reports Trahan’s estimate that, if the bill passed in late 2026, its provisions would expire in December 2029. The draft text was still being revised when the guide was published, and its final wording is not known.

Pre-emption is the key fight. Several states already have frontier AI laws, and the NHIMG guide compares the draft with them. A federal law that overrides those rules would change which requirements AI companies must follow. The guide does not say what state officials think of the draft. Public reaction from them could not be confirmed.

Advertisement

A ban on superhuman AI

Other lawmakers want to go much further. According to an essay published on 12 November by Resilience, Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) plan to introduce legislation that would ban the development of artificial superintelligence. As the essay describes it, the bill would permanently ban superhuman AI, pause advanced AI development and set up a federal agency to regulate the technology.

The essay describes this bill as future legislation, not as a bill already filed. No bill text or number was available in the sources reviewed for this article.

The same essay goes on to propose its own model bill, the “Limits to AI Act”. Under that proposal:

Advertisement
  • The Federal Energy Regulatory Commission would take over all data centres used for advanced AI.
  • The Federal Communications Commission would decide which AI uses count as “limited AI” that private companies may offer and which count as “advanced AI” reserved for the federal government. It would also decide which systems qualify as artificial general intelligence or superintelligence.
  • AI data centres run by the federal government would have to use renewable energy, have a water conservation plan and limit noise.
  • Breaking the bans on artificial general intelligence and on private AI development would be a crime.

The essay does not name a congressional sponsor for this model bill. For now it is an advocacy proposal, not legislation that has been filed.

How the US compares

Other jurisdictions already have laws in force. According to iTechGuides, an artificial intelligence regulation register listed three comprehensive AI laws as of 2 November 2026: those of the European Union, South Korea and Vietnam.

The EU’s AI Act has been delayed only in part. The overview says the main requirements for high-risk systems now apply later. The bans on prohibited practices, the duties for general-purpose AI models and the Article 50 transparency rules were not postponed and were in force as of 10 November. The Mean CEO roundup cites a Council on Foreign Relations guide saying the Act entered its enforcement phase in summer 2026.

Advertisement

In the US, the overview describes federal regulation as existing laws plus a growing patchwork of state requirements. It says registration under one state’s frontier AI rules is reported to begin in November 2026. It also reports a 72-hour window for reporting incidents, with penalties of up to $1 million for a first violation and $3 million for a repeat. The overview presents those details as reported rather than confirmed.

What would need to happen next

For the Great American AI Act, the first step is formal introduction. That would give the draft a bill number and published text. Committee hearings and a markup would follow, and at each stage the frontier duties and the pre-emption clause could change. Its sponsors have already said they expect to negotiate.

The proposed ban on superintelligence would also have to be introduced first. Until then, it is not clear how it would define superhuman AI, which agency would enforce it or how a pause on advanced development would work. A Congress that has not passed any comprehensive AI statute would then have to decide between a federal framework that limits state rules and a federal ban on the most advanced systems.

Advertisement

Companies building or using AI in the US will keep following state laws and existing federal rules for now. Firms that sell in Europe already face EU obligations that are in force.

Sources