Federal vs State AI Preemption: Why You Still Comply With State AI Law Right Now
Bindingness: Non-Binding Framework · Scope: Federal · Multi-sector
Washington has launched a litigation program to flatten the state AI patchwork. It has not succeeded. Here is why your compliance program must not change until it does, and the posture that survives either outcome.
Part of AI Regulation Tracker, our running brief on the laws reshaping AI in professional work.
There are two stories running at once, and confusing them is the expensive mistake. The first story is loud: the administration has declared war on the state AI patchwork, stood up a Department of Justice task force whose only job is to sue states, and intervened in court to oppose the country's most prominent AI law. The second story is quiet and is the one that governs your week: not one of those state laws has actually been preempted. The fight over who wins is real and could run for years. The rules you are bound by today did not move.
This briefing is for the general counsel, the compliance officer, and the professional in any field who is being asked some version of the same question right now: the feds are challenging these laws, so can we ease off? The short answer is no, and the reason is mechanical, not political. This piece explains the federal push in plain terms, why it does not yet change your obligations, and the multistate posture that holds up whether preemption succeeds or fails. It builds on our briefing on the 2026 federal AI executive order, which covers the order itself; here we focus on the litigation effort behind it and what it means for your compliance program.
Key takeaways
- The December 11, 2025 executive order directed the Attorney General to stand up an AI Litigation Task Force to challenge state AI laws on interstate-commerce and preemption grounds. The DOJ created it on January 9, 2026.
- An executive order has no preemptive force on its own. It is not a statute. Only Congress, a valid federal regulation issued under congressional authority, or a court can actually override a state law.
- The task force is already in court. The DOJ moved to intervene against Colorado's AI law in April 2026, the first time it has joined a challenge to a state AI statute, and enforcement of that law was suspended while the case proceeds.
- State laws keep passing. Connecticut enacted a sweeping AI law in June 2026. The patchwork grows even as Washington tries to flatten it.
- The durable posture is to comply with the strictest state you touch and keep that standard in place. It satisfies every lower bar and survives whichever way the preemption fight breaks.
What the federal push actually is
On December 11, 2025, the White House signed an executive order titled "Ensuring a National Policy Framework for Artificial Intelligence." It sets a federal goal of one national rulebook for AI and treats the growing thicket of state laws as an obstacle to interstate commerce. Three instructions inside it matter. It directs the Attorney General to establish, within 30 days, an AI Litigation Task Force whose sole job is to challenge state AI laws the administration views as unconstitutional or preempted, expressly including the argument that such laws unlawfully regulate interstate commerce. It directs the Commerce Department to publish, within 90 days, an evaluation flagging state laws it considers onerous and to refer them to the task force. And it leans on federal funding, conditioning certain broadband money on whether states pause or align their AI enforcement.
The DOJ stood up the task force on January 9, 2026, with a memorandum from the Attorney General to all department employees framing the mission as freeing AI companies from cumbersome regulation. In March 2026 the White House released a set of legislative recommendations asking Congress to enact a uniform federal framework that would preempt conflicting state laws, with carve-outs for child safety, data-center infrastructure, and state procurement and government use of AI. A second executive order on June 2, 2026, titled "Promoting Advanced Artificial Intelligence Innovation and Security," widened the cybersecurity and innovation agenda and reinforced the same preemption strategy, again pressing the funding lever and the litigation track. Taken together, this is a coordinated, three-pronged effort: sue the states, ask Congress to legislate them away, and squeeze them with federal dollars.
Why none of it changes your obligations yet
Here is the part that should govern your decisions. An executive order is a direction to the executive branch. It is not a law passed by Congress, and it is not a regulation issued under a law Congress passed. On its own it has no power to strike down a duly enacted state statute. Preemption, when it happens, comes from one of three places: Congress passes a federal law that displaces the state rule, a federal agency issues a valid regulation with preemptive effect under authority Congress delegated, or a court rules that a specific state law is unconstitutional or preempted. Until one of those things happens for a law you are subject to, that law is in force and you are bound by it.
A government that announces it will challenge a law has not repealed the law. The gap between the announcement and the outcome is measured in years, and you live inside it.
Congress has repeatedly declined to pass broad federal preemption of state AI laws. The litigation track is live but unresolved. The clearest example is Colorado. xAI sued to challenge Colorado's AI law in early April 2026, the DOJ moved to intervene on April 24, 2026 to oppose the same law, and on April 27 the court temporarily suspended enforcement of that statute while the case proceeds. Even there, Colorado did not surrender the field: in May 2026 the legislature passed Senate Bill 26-189, which repealed and reenacted its AI law in a narrower form, replacing the heaviest obligations with a set of operational duties around automated consequential decisions and pushing the effective timeline to January 1, 2027. The pattern is instructive. One marquee law got tangled in litigation and rewritten; the obligation did not vanish, it changed shape. Meanwhile other states kept legislating, with Connecticut enacting a broad AI law covering employment, healthcare, and online safety in June 2026. The patchwork is contested, not gone.
Challenging a law is not the same as winning
The task force can file suit, intervene, and argue preemption. It cannot erase a state statute by filing. A challenge in progress is a risk to track, not a reason to relax. If you stand down on the strength of a suit that has not been decided, you are betting your compliance posture on a verdict no one has reached.
The trap: reading the federal posture as permission
The dangerous move is to hear "the administration is suing to kill these laws" and quietly loosen your controls. That single decision stacks three bets on top of each other. You are betting the challenge succeeds, betting it succeeds soon, and betting that no state agency or private plaintiff acts against you in the meantime. Each bet can fail independently, and the downside is a regulatory action or a lawsuit under a law that was on the books the entire time. The cost of holding your controls in place while the lawyers fight is low. The cost of guessing wrong about preemption is not.
This matters across professions, not just for the AI labs in the headlines. If you are using AI in hiring or promotion decisions, in consumer-facing lending or insurance, in healthcare workflows, or in any consequential decision about a person, you are touching exactly the conduct these state laws govern. The federal challenge does not reach down and immunize your specific use. It contests laws in court while your obligations under them continue. Compliance leaders who treat the EO as a green light are misreading a fight as a finish line.
Federal vs state status: where each lever actually stands
| Federal lever | Status (June 2026) | Effect on your duty today |
|---|---|---|
| Dec 11, 2025 EO (national framework) | In effect as policy and agency direction | None directly. It cannot repeal a state law by itself. |
| DOJ AI Litigation Task Force | Created Jan 9, 2026; litigating | None until a court rules. A pending suit is a risk to monitor, not a repeal. |
| Commerce evaluation of state laws | Identifying and referring laws | Signals targets. Does not change a referred law's force. |
| March 2026 legislative recommendations | Recommendations only; not enacted | None. Congress has repeatedly declined broad preemption. |
| June 2, 2026 EO (innovation and security) | In effect; reinforces strategy | None directly on state-law compliance. |
| State AI laws (CO, CT, IL, CA, others) | In force or on timelines; still being enacted | Binding. You comply until Congress or a court says otherwise. |
The posture that survives either outcome
You do not need to predict who wins the preemption fight to build the right program. The clean answer to a contested, moving patchwork is to stop designing a separate process for each rule and instead build one program to the strictest obligation you face anywhere you operate. If your highest-bar state requires user notice, a documented review of consequential AI decisions, recordkeeping, and a named human owner, build all of that into one company-wide standard and run it everywhere. A program set to the high-water mark is automatically compliant with every lower bar, does not have to be rebuilt if a single law is struck down, and does not collapse if a new state joins the list next quarter. It is the one posture that holds whether preemption succeeds, fails, or stalls.
Multistate AI-Compliance Posture Decision Guide
- Map your footprint to obligations. List every state where you operate or make AI-touched decisions about people, and the specific AI obligation each one imposes today. You cannot comply with the strictest state until you know which states you touch.
- Set one company-wide standard to the high-water mark. Adopt the strictest single version of notice, human review of consequential decisions, recordkeeping, and disclosure, and apply it everywhere rather than maintaining a different process per state.
- Treat a federal challenge as a watch item, not a release. When a law you rely on is challenged or its enforcement is suspended, log it, keep your controls running, and reassess only when a court actually rules or Congress actually acts.
- Name a human owner of record for every covered decision. Whatever rulebook ends up applying, you must be able to show a named person owned the outcome, not the model.
- Keep a living legal-status matrix. One row per state: the obligation, its current status, any pending challenge, and the strictest version you have adopted. Refresh it on a fixed cadence because both the laws and the litigation are moving.
- Decide changes on rulings, not on rhetoric. Relax a control only when a binding event removes the underlying obligation, never on the strength of an executive order, a lawsuit in progress, or a press release.
Building and maintaining that matrix is exactly the kind of work an AI assistant handles well, and it is also where the discipline shows. Claude can draft a plain-language compliance matrix across your states, summarize each obligation and its litigation status, and keep the document current as rules and cases move faster than any person can track by hand. What it cannot do is decide how preemption will break, tell you which law governs a specific decision, or carry the liability if your process produces a bad outcome. The tool builds the map. A named human in legal or compliance confirms each entry against current guidance and owns the result. That separation, fast machine drafting plus human judgment on the line, is the whole skill.
The next ninety days are a monitoring problem, not a standing-down problem
Expect more task-force filings, more Commerce referrals, possibly a court ruling on Colorado, and more states enacting laws of their own. None of that is a reason to ease controls in advance. Track the events, keep your high-water-mark standard running, and move only when a binding ruling or statute actually changes the obligation.
How to use AI through the uncertainty without getting it wrong
The lesson here is not to fear the tool or freeze the program. Professionals who refuse AI while the rules settle will lose ground to those who use it to stay current and well documented. The lesson is that the federal-versus-state fight changes the speed at which you must track obligations and changes nothing about the duty to actually meet them. Using AI to build and maintain a defensible compliance posture, while keeping the legal judgment with counsel and a named owner, is a skill you build deliberately, not improvise under deadline. That is precisely how we teach professionals to work with Claude: take the leverage on drafting and tracking, keep the judgment and the accountability human.
Frequently asked questions
Does the federal executive order cancel state AI laws?
No. The December 11, 2025 and June 2, 2026 executive orders direct federal agencies to challenge and try to preempt state AI laws, but an executive order has no power to strike down a state statute on its own. It is not a law passed by Congress or a regulation issued under one. Until Congress enacts preemption or a court rules a specific law unconstitutional or preempted, the state law remains in force.
What is the DOJ AI Litigation Task Force?
It is a Department of Justice unit created on January 9, 2026 under the December 2025 executive order, with the sole job of challenging state AI laws the administration considers unconstitutional or preempted, including on the theory that they unlawfully regulate interstate commerce. It is already active: the DOJ moved to intervene against Colorado's AI law in April 2026, its first intervention in a state-AI-law challenge.
If a state law I rely on is being challenged, can I stop complying?
No. A pending challenge, or even a temporary suspension of one law's enforcement during litigation, does not remove your obligations under other state laws or guarantee any particular outcome. Relax a control only when a binding event, a court ruling or a statute, actually removes the underlying duty. Standing down on the strength of a lawsuit in progress bets your compliance posture on a verdict no one has reached.
How should a multistate organization operate while preemption is unsettled?
Comply with the strictest state you touch and run that standard everywhere. Build one company-wide program set to the high-water mark for notice, human review of consequential decisions, recordkeeping, and disclosure. A program built to the strictest requirement is automatically compliant with the lower bars, does not need rebuilding if one law is struck down, and survives whichever way the preemption fight breaks.
Did Colorado repeal its AI law because of the federal pressure?
Colorado did not abandon AI regulation. After its original law drew a federal callout, a private lawsuit, a DOJ intervention, and a court-ordered enforcement suspension, the legislature passed Senate Bill 26-189 in May 2026, which repealed and reenacted the law in a narrower form with new operational duties and a January 1, 2027 effective date. The obligation changed shape rather than disappearing, which is the pattern to expect elsewhere.
Related reading: the 2026 federal AI executive order explained, the AI employment-law patchwork for HR, and what Colorado's AI hiring law means for compliance.
Sources: White House, "Ensuring a National Policy Framework for Artificial Intelligence" executive order (December 11, 2025) and "Promoting Advanced Artificial Intelligence Innovation and Security" executive order (June 2, 2026); CBS News, "DOJ creates task force to challenge state AI regulations" (January 2026); Ropes and Gray, "Trump Attempts to Preempt State AI Regulation Through Executive Order" (December 2025) and "Examining the Landscape and Limitations of the Federal Push to Override State AI Regulation" (March 2026); Latham and Watkins, "AI Executive Order Targets State Laws and Seeks Uniform Federal Standards"; Paul Hastings, "President Trump Signs Executive Order Challenging State AI Laws"; Norton Rose Fulbright, "X.AI sues, DOJ intervenes, enforcement of Colorado AI Act suspended"; Crowell and Moring, "Colorado Hits Reset on AI Regulation: SB 26-189 Repeals and Reenacts the Colorado AI Act." This briefing is general information, not legal advice. AI laws and the litigation over them are changing quickly; confirm against current authority and counsel before relying on this page.