New White House Guidance Sidesteps Federal Preemption of State AI Laws

Alexis Keenan Published March 25, 2026

California’s AI laws were indirectly targeted for federal preemption on Friday in a new “National Policy Framework” penned by the Trump administration.

However, in a surprise to stakeholders, the new policy steered clear of singling out any state AI statutes and omitted more detailed preemption language laid out in a December executive order.

“This is a lot of talk with not a lot of action behind it,” Alia Luria, a former software engineer turned data privacy and technology compliance lawyer, said about the new policy. “[It’s] even less of a stance than I expected it to be.”

The four-page policy sets forth “legislative recommendations” for the roles that the administration says states and the federal government should play in regulating AI.

“States should not be permitted to regulate AI development, because it is an inherently interstate phenomenon with key foreign policy and national security implications,” the policy states, echoing President Trump’s December executive order. “Preemption must ensure that State laws do not govern areas better suited to the federal government or act contrary to the US’ national strategy to achieve global AI dominance.”

Stakeholders had expected the policy to explicitly call out AI regulations, such as California’s AI Transparency Act (partly delayed until Aug. 2, 2026) and AI Training Data Act, as well as Colorado’s delayed AI Act and laws governing the technology in New York, Georgia, Illinois, Iowa and Maryland.

Experts said New York’s RAISE Act was a suspected target, as were a growing body of state laws addressing deepfakes, chatbot disclosures and automated decision-making in employment.

Instead, the policy made general statements recommending invalidation of “burdensome” state laws. In a departure from the executive order, the framework focused more on lawmakers passing preemptive federal statutes than on federal agency challenges to state laws.

“Congress should preempt state AI laws that impose undue burdens to ensure a minimally burdensome national standard consistent with these recommendations, not fifty discordant ones,” the policy said.

Luria said the published language still indicates that the administration is intent on pursuing enforcement of allegedly unlawful state statues through the FTC, because the same document recommends that Congress take a hands-off enforcement approach.

“Congress should not create any new federal rulemaking body to regulate AI, and should instead support development and deployment of sector-specific AI applications through existing regulatory bodies with subject matter expertise and through industry-led standards,” the blueprint said.

Luria said the guidance reads as “kicking the can down the road…with a dash of trying to subtly threaten AI companies regarding any quashing of what the administration considers protected speech.”

Congress, the document said, should prevent the government from coercing AI providers to ban, compel or alter content based on partisan or ideological agendas. Meanwhile, it added, the government should prevent AI systems from silencing or censoring political expression.

On the other hand, President Trump’s executive order – “Ensuring the United States Remains the World Leader in Artificial Intelligence” – gave detailed directions to multiple federal agencies to evaluate state AI laws for conflicts with federal law, to publish the evaluations and to challenge the laws identified as problematic.

Unlawful state AI statutes, the order said, could include those that unconstitutionally regulate interstate commerce, violate the First Amendment or are preempted by existing federal regulations.

The order instructed U.S. Attorney General Pam Bondi to establish a new AI Litigation Task Force to challenge unlawful state statutes, and instructed U.S. Secretary of Commerce Howard Lutnick to identify and refer to the task force “onerous” state AI laws that conflict with national policy.

The Commerce Department evaluation was to focus on state laws that require AI models “to alter their truthful outputs,” and those that “compel AI developers or deployers to disclose or report information in a manner that would violate the First Amendment,” or other constitutional provision.

“Onerous” state laws, the order said, would trigger notices to states of their ineligibility to receive funds from the Broadband Equity, Access, and Deployment (BEAD) Program, a $42.45 billion grant program to help states connect Americans to high-speed internet.

The order also provided a mandate for the U.S. Federal Trade Commission.

It directed FTC Commissioner Andrew Ferguson to issue a policy statement explaining how the FTC Act’s prohibition of unfair and deceptive acts and practices affecting commerce applies to AI models, and to identify circumstances under which state laws that “require alterations” to an AI model’s truthful outputs are preempted by the Act.

Federal Communication Commission Chairman Brendan Carr, for his part, was ordered to initiate proceedings for a national AI disclosure standard.

However, no policy statements or evaluations were included or referenced in the published framework, perhaps exposing the significant hurdles inherent in overturning state-level AI laws.

Federal Preemption v. State Rights

“Without [federal] legislation, the administration is essentially running an influence campaign backed by litigation threats and funding conditions,” said Ross Guberman, founder and CEO of BriefCatch, a legal writing software used by more than 50 AmLaw 200 firms and federal courts. The approach, he said, can slow the pace of state regulation but probably can’t stop it.

Trademark and intellectual property law expert Jessica Eaves Mathews said the concept of states’ rights, secured by the 10th Amendment, is so revered in the U.S. that it would be a huge challenge to get the needed majority of Congress members to preempt states from regulating AI within their own borders.

“It’s sacred to most lawmakers, and more so to the GOP, even than Democrats,” she said. And another hurdle stands in the way even if Congress agreed on some level of AI regulation: “To get them to pass legislation is one thing, but if Trump vetoes or doesn’t like it, overcoming a veto would be even harder, because majorities are razor-thin.”

One piece of legislation known as the “TRUMP AMERICA AI Act“, led by Sen. Marsha Blackburn (R-Tenn.), seeks to preempt state AI regulation by replacing the “patchwork” of rules with a single federal rulebook.

A counter-proposal from a coalition of Democrats, known as the “GUARDRAILS Act” – introduced by Rep. Don Beyer (D-Va.) and Sen. Brian Schatz (D-Hawaii) – seeks to repeal President Trump’s December executive order.

Guberman said he had been closely watching for how the Commerce Department and the FTC would define “onerous” and “deceptive.” His clients include major law firms and courts, which he said are looking for regulatory clarity.

For BriefCatch, Guberman said, the main question is whether the company’s products fall within or outside the crosshairs of a federal preemption push. For the company’s clients, he said, their main question is how to remain compliant state to state.

“Right now, they’re navigating a growing patchwork of state rules — not just about how AI is built, but about how lawyers and judges can use it,” Guberman said. “Anything that simplifies that landscape helps them adopt tools like ours more confidently.”

Funding Enforcement Mechanisms

Without more clarity, Guberman and other stakeholders said, the president’s order and new guidance are most influential as tools to pressure state regulators.

“I think Trump understands he’s probably not going to get Congress to preempt the state’s ability to regulate AI for interstate activities,” Eaves Mathews said. “Now, there could be pressure brought to bear. There could be intimidation and funding issues.”

Guberman described the BEAD funding ultimatum as “a meaningful financial incentive” that will make states think twice before enacting aggressive AI regulation.

He noted recent bipartisan sentiment against usurping state power. In July, U.S. senators rejected 99-1 a 10-year moratorium on state AI laws proposed in the president’s omnibus spending package, the “One Big, Beautiful, Bill Act.”

In November, 36 state attorneys general joined a letter opposing executive branch attempts to place limits on state AI regulation and enforcement.

“Even the Heritage Foundation is coming down on the side of states should be able to regulate this,” Eaves Mathews said about the conservative think tank, which said states should be left alone to protect children and workers.

The rejected moratorium in the president’s omnibus package, the Heritage Foundation’s technology researcher said, threatened to “override legitimate state efforts to curb Big Tech’s worst abuses—with no federal safeguards to replace them.”

In addition to the BEAD funding ultimatum, the president’s order directed executive agencies to investigate their authority to condition other federal funding grants on a state’s AI regulations – or lack thereof.

“Truthful Outputs” and Algorithmic Bias

One legal argument from the White House that could be viable, although it’s yet to be tested, is that certain state anti-discrimination laws governing AI may force developers or providers to produce “deceptive” or untruthful results.

“State laws are increasingly responsible for requiring entities to embed ideological bias within models,” the executive order said, singling out Colorado’s AI Act, which limits “high-risk” AI applications by imposing bias testing and impact assessment obligations.

“The administration’s theory is that state laws requiring AI models to mitigate bias effectively force developers to produce less accurate outputs, making those outputs ‘deceptive’ under the FTC Act,” Guberman explained. “That’s a novel and untested legal theory.”

“That’s an extremely new way of looking at this,” Luria said, though she said she views the administration’s policy primarily as a means of chilling state regulation.

“The precedent for suing states for deceptive trade practices is not there,” she said. “So the question is: is this to try and control the purse strings? Or is it to try and get rid of as many AI disclosure laws as possible, which are the low-hanging fruit?”

Pressure from the White House and technology lobbies may be intensifying.

Back in August, Colorado state lawmakers delayed the AI Act’s original Feb. 1, 2026 effective date to June 30, 2026, to buy more time for compromise with technology heavyweights that rejected provisions that risked more liability based on how a local company or user might deploy their AI tools. The opponents also rejected the law’s prohibition of so-called “algorithmic discrimination” and its European Union model of increased liability for “high-risk” AI systems.

With no compromise yet in place and the White House’s policy framework now in play, Colorado Gov. Jared Polis on March 17 announced a working group effort to “repeal and replace” the disputed provisions. The group proposed a bill to amend the Act, and further refinements to limit developer liability were proposed on March 23.

Polis, who noted at the time he signed the legislation that he did so reluctantly, also supported the failed House GOP proposal to ban state laws restricting AI.

Innovation and Geopolitical Competition

As with prior emerging technologies, the administration’s policy exposes tension between the federal government’s desire for a unified national standard that fosters innovation, and the states’ authority to regulate for the safety of their citizens.

“The United States must lead the world in AI by removing barriers to innovation, accelerating deployment of AI applications across sectors, and ensuring broad access to the testing environments needed to build world-class AI systems,” the framework said.

In the 1990s, Congress passed the Internet Tax Freedom Act, placing a “moratorium” on states’ taxes on internet access. States argued that the Act starved them of vital tax revenue, while the government said thousands of distinct taxes would stifle a nascent technology.

“We must not strangle the infant Internet in its crib with a patchwork of 30,000 different tax jurisdictions,” U.S. Rep. Christopher Cox said during a congressional debate at the time.

Eaves Mathews said that “if you look historically at what’s happened in this country, there’s an overriding belief that regulation is antithetical to innovation.”

“The big concern, I think, that people in Congress have, is falling behind countries like China,” she continued. “Because we’re in basically a new arms race.” If it doesn’t let tech develop unfettered, the thinking goes, the U.S. will end up on the losing end.

Although the White House’s new framework suggests that Congress steer clear of creating new rulemaking bodies to regulate AI, it does encourage lawmakers to create “regulatory sandboxes” for AI applications to help “unleash American ingenuity.”

Dhruv Patel, co-founder and CEO of the local grant procurement software company Syncurrent, said even federal preemption could work against the administration’s goals.

“There’s a lot of learning that needs to happen with AI development,” he said. “I think sweeping federal regulation out of the gate, at this stage, could be a bad idea.”

Instead, he said, the federal government could look to its existing non-regulatory framework for design, development, use and governance of AI technologies created by the Commerce Department’s National Institute of Standards and Technology (NIST).

The voluntary framework is intended to maximize enhanced economic security, competitiveness and quality of life through research.

“The FTC and the Department of Commerce already have authority in holding companies accountable for their business practices,” Patel said. “Using NIST as a best practice gives the federal government a baseline to measure companies who utilize AI.”

Protecting Children, Public Safety

Although the federal policy aims to temper state regulation, it explicitly supports states’ rights to regulate AI to protect children, prevent fraud and protect consumers.

Also off limits for preemption, the policy said, are state laws governing a state’s own use of AI – for example, to provide services such as law enforcement and public education.

California and 44 other states have enacted AI laws protecting children primarily through the criminalization of AI-generated child sexual abuse material (CSAM). “Children’s protection and privacy is a big area where California is pushing forward,” Luria said.

However, the guidance also offers Big Tech a bonus by encouraging Congress to shift liability for harms to children from technology developers and providers to parents.

“The administration is calling on Congress to give parents tools…such as account controls to protect their children’s privacy and manage their device use,” the guidance said.

Impact on the Legal Profession and Judiciary

The administration’s stance could also influence the way lawyers use AI, and how courts regulate the technology.

“Right now, the judicial system is developing its own regulatory ecosystem for AI, largely independent of what Congress or the White House does,” Guberman said.

Dozens of federal judges have issued standing orders requiring lawyers to disclose when they’ve used generative AI, he explained. And courts are actively handling disputes over the new technology, such as large language model “hallucinations” in court documents, and claims of intellectual property infringement.

The president’s executive order primarily targets how AI models are built, rather than how AI tools are used in specific contexts, such as within the judicial system. However, the new guidance did weigh in on unsettled litigation, saying the administration believes that training AI models on copyrighted material does not violate copyright laws.

However, it said that “Congress should not take any actions that would impact the judiciary’s resolution of whether training on copyrighted material constitutes fair use.”

Litigants, business owners, developers, state regulators, and other stakeholders will have to wait to learn which levers the administration will ultimately push to limit more than 1,500 AI-related bills introduced in 45 states as of March 2026.

“It’s unnerving, I think, for a lot of businesses to know what to do right now,” Eaves Mathews said.

Luria expects the administration to first test the FTC’s luck at challenging state laws. Eaves Mathews said the clash is going to come down to negotiations, pressure and influence more than anything else.

“I think the idea that we may see some meaningful preemption, where the federal government is able to control the regulatory regime in in the U.S., is really unlikely,” she said. “But it doesn’t mean Trump’s not going to try.”

© The Regents of the University of California, 2026.