Summary: The December 11, 2025 executive order sets March 11, 2026 deadlines for Commerce to identify state AI laws it considers onerous and for the FTC chair to publish a preemption policy statement. The order does not itself preempt or invalidate state laws, and roughly $21 billion in remaining BEAD funds is described as leverage. The article says existing state frameworks remain enforceable absent court action.
The 90-day deadlines inside the executive order President Trump signed on December 11, 2025 — “Ensuring a National Policy Framework for Artificial Intelligence” — come due on March 11, 2026. Commerce must by then publish its evaluation of state AI laws and single out the ones it judges “onerous,” while the FTC chairman has to put out a policy statement on when those state laws are preempted by federal law. The White House page’s URL is blunter than the title: “eliminating state law obstruction” of national AI policy.
The significance: this is the first time the federal government has turned its squeeze on state-level AI regulation into an administrative process with a deadline and money attached. For teams shipping AI products into the US market, the compliance map may be redrawn — or the whole question may simply sit in litigation for years.
Machinery First, Repeal Not Yet: What the December Order Does
Its stated policy goal is “global AI dominance through a minimally burdensome national policy framework for AI.” As a piece of law, though, the order is restrained: it does not itself preempt, repeal, or invalidate any state law. What it actually does is put federal agency actions into a fixed sequence.
One statute is called out by name: Colorado’s AI Act (SB 24-205), which obliges developers of high-risk AI systems to take reasonable care against “algorithmic discrimination” and takes effect June 30, 2026. The administration’s complaint is that duties like this could force models to alter their truthful outputs. The companion move followed on January 9, 2026, when Attorney General Pam Bondi created the DOJ’s AI Litigation Task Force, charged with challenging state AI laws on grounds including unconstitutional burdens on interstate commerce and federal preemption. As of early March, not one lawsuit had come out of it — partly because Commerce had to finish its identification first.
The Two Documents That Must Land by March 11
- Commerce (Section 4): an assessment naming the state laws it considers “onerous” — above all those that “require AI models to alter their truthful outputs,” along with compelled disclosures that raise First Amendment concerns.
- FTC (Section 7): the chairman must lay out when laws demanding altered outputs are preempted by Section 5 of the FTC Act, the provision covering unfair or deceptive practices. No court has tested this theory: state laws that mandate output adjustments to mitigate bias could, by its logic, push developers into producing “deceptive” results under federal law.
Taken together, the pair amounts to a target list with federal backing plus a legal argument — the list is for the DOJ to litigate, the theory for the courts to decide.
Leverage Worth $21 Billion
Section 5 puts an instrument in the administration’s hands that never has to enter a courtroom: a Commerce policy notice tying roughly $21 billion in remaining BEAD (Broadband Equity, Access, and Deployment) nondeployment funds to states not maintaining such laws.
This is federal money deployed as leverage around the legislative process — with no act of Congress and no court order needed. States short on cash will feel it as real pressure; anyone tracking the legal fight can read it as the clearest measure of how much further the administration intends to push.
Which State Laws Could Make the Risk List
Exactly one law gets named in the order itself: Colorado’s SB 24-205. Law-firm inventories point to a much wider pool of candidates: California SB 53 (frontier model transparency), California AB 2013 (training data transparency), New York’s RAISE Act (signed December 19, 2025), and the expanding run of state laws on deepfakes, chatbot disclosures, and automated decision-making in employment.
Breadth is the indicator to watch. A short list aimed at a few omnibus statutes means surgical strikes; one stretching to hundreds of narrower laws means total war. And the longer the list grows, the bigger the litigation queue — and the uncertainty riding along with it.
The Practical Read for Engineering and Legal Teams
Begin with what has not changed: no law has been struck down. California’s, Texas’s, and other state AI frameworks already in effect “remain fully enforceable absent court action.” For any of them to actually fall, the DOJ has to sue and a court has to grant an injunction — a timeline measured in months to years.
Baker Botts’ advice to companies is level-headed: keep complying; map which obligations could face federal challenge; prepare a contingency plan for each; and track three things — how broad the evaluation turns out, the legal theory the FTC runs with, and the state the DOJ comes after first. Translated for product teams, it is simpler: for AI products delivered across state lines, the strictest state law still sets your compliance floor, and what this federal offensive adds is uncertainty, not clarity. The easiest trap of 2026 is treating state-law obligations as tech debt that repeal will quickly clear away.