The Fight Over Who Gets to Regulate AI in America Is Getting Sharper

The Fight Over Who Gets to Regulate AI in America Is Getting Sharper

Washington, or individual states? Who is better able to regulate artificial intelligence and its many potential abuses? This question has resurfaced in a particularly heated debate in the United States of this year, with politicians on both sides of the pond wrestling with this thorny problem.

Here's the situation at the time being, and why the battle has such resonance beyond just the field of technology.

Summary of background and what happened so far

All of this began on December 11, 2025, when President Trump signed an executive order to have the federal government take responsibility for regulating AI and limit the number of state laws interfering with this.

On this same day, the President ordered the Attorney General to sue states whose laws were judged too restrictive to the development of AI.

And, the Justice Department has announced in January 2026 the creation of a task force to sue states with "unduly burdensome" measures on artificial intelligence.

Trump's executive order gives the Justice Department the leverage to use "condition precedent" to withdraw infrastructure funds or other forms of federal assistance to states which maintain laws judged "inconsistent or incompatible" with the national standards relating to AI.

The framework document is a non-binding recommendation to Congress as to what national AI legislation should contain.

This text, composed of 4 pages, details recommendations which should, according to the White House, guide Congress in its discussions on AI.

The document is structured as an enumeration of topics on which action should be taken.

These are: protection of children on social media, electricity costs of AI computing, intellectual property, free speech, national security (concerning so-called 'frontier' or advanced AI models), and lastly workforce development.

For the issue of AI firms' rights to use copyright-protected material for training models, the text describes the view of the administration, which states that it considers this practice lawful, although the document also recognises this as a grey area in current law, and recommends leaving this question to courts.

In relation to small businesses adopting AI technologies, the document recommends that the federal authorities take action to facilitate their access to AI and make it easier for them to invest in AI, perhaps through grants or other financial tools.

The most far-reaching argument advanced in support of pre-emption in this document concerns the threat to competitiveness posed to the American tech sector by the EU and China unless the United States adopts a unified national framework.

One way or another, this approach provides for lighter federal regulation than the state law alternatives which are available to states currently.

At the same time, the main arguments made by proponents of this view also apply to state laws currently in force or debated in states.

This is because, until now, state laws were the only form of regulation in the United States, whereas federal laws were not implemented.

And, it is not necessarily the case that state AI laws are redundant or ineffective.

According to Brad Carson, president of the Algorithmic Responsibility Institute, one of the main proponents of federal preemption, this represents an attempt, in his words, to "turn the current floor on state AI legislation into a federal ceiling".

On the other hand, it is argued that the burden placed on companies by having to comply with numerous, possibly conflicting sets of state regulations is excessive and unproductive in terms of accelerating the production of generative AI systems; this is the view advocated by major tech companies currently.

According to this argument, there is no need for state laws either, because this multiplicity of regulations ends up being a burden on companies and does not make society any safer.

The balance of power between the states and the Trump administration

Despite the controversy it has aroused, the text has received support from a number of influential Republicans in addition to the President, such as Marco Rubio, Rand Paul, Ted Cruz and Cory Gardner.

On the other side, in addition to pro-technology lobbies and major industry players, consumer defence organisations also oppose it. This is because the state laws in force at the moment provide a basic level of protection which cannot currently be superseded by federal law.

However, the situation is complex in that the most important laws on AI are those which have been passed or are currently being considered at the state level, rather than in Congress.

A MultiState analysis of March 2026 shows that legislators in 45 of the 50 states have presented a total of 1,561 bills related to AI since January 1, 2026, which already exceeds the total number of bills relating to AI introduced in the United States in 2024.

It is difficult to assess the effectiveness of state laws currently in force in dealing with the various challenges of AI regulation.

Thus, in March 2026, the White House presented what it calls its "National Policy Framework for Artificial Intelligence"; this consists of a set of recommendations to Congress regarding the regulation of AI technologies.

The document describes, among other things, the main areas to be regulated.

It is useful to note that, in most cases, these cover a number of problems at once.

As such, the White House statement presents a number of recommendations to Congress concerning the regulation of the use of AI.

Some of these recommendations were previously made by state legislators or expressed by various industry players.

At the moment, Congress does not seem to have adopted any bills which implement any of these recommendations.

On the contrary: in 2026, Congress has so far rejected proposals to adopt a comprehensive national AI strategy, including a proposal earlier in the year to impose a temporary moratorium on state laws relating to AI.

The only substantial piece of national legislation at the moment is the TAKE IT DOWN Act adopted in May 2025 and currently in force as of May 2026; it concerns the obligation of platforms to remove non-consensual sexual imagery, including images generated by AI, and is enforced by the FTC.

There is also another proposed federal law, sometimes called the Obernolte-Trahan framework.

This project has received favourable mention from major industry players in seeking clarity on the regulation of AI, but it has been criticised by consumer advocacy groups on the same "ceiling versus floor" issue, which the text itself has not yet adopted as official legislation, and thus it is still a draft currently awaiting formal presentation and vote in Congress.

Bottom Line

At the moment, the basic situation is that state laws are in force and can be applied immediately.

The dispute between the states and the federal government concerning the regulation of AI has not so far resulted in the abrogation of these state laws.

The outcome of this discussion in 2026 will have implications on the development and deployment of generative AI within the United States, which both sides of the debate consider extremely important.