Key Takeaways
- Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," signed December 11, 2025, directed the Attorney General to stand up an AI Litigation Task Force within 30 days to challenge state AI laws on dormant-commerce-clause, preemption, and other constitutional grounds. The Department of Justice formalized the task force by memorandum on January 9, 2026.
- The order does not touch state laws on child safety, AI compute and data center infrastructure, or state government procurement and use of AI. Those carve-outs are written directly into Section 8(b) of the order.
- The task force's first real-world test arrived faster than most forecasts assumed: xAI sued Colorado on April 9, 2026 over its AI Act, DOJ intervened on April 24, and a federal court stayed enforcement of the law on April 27, just weeks before its scheduled effective date.
- Congress has twice declined to do by statute what the executive order tries to do by litigation and funding pressure: the Senate stripped a ten-year AI moratorium from the 2025 reconciliation bill 99-1, and a second attempt to preempt state AI laws through the FY2026 NDAA was dropped before passage.
- None of this has invalidated a single state AI law yet. Law firms tracking the litigation, including Paul Hastings and Latham & Watkins, are telling clients to keep complying with state AI laws as written, because an executive order cannot overturn a statute. Only Congress or a court can do that, and neither has.
If your AI governance roadmap has a line item that says "pause pending the executive order," that line is the wrong call. Nothing about Executive Order 14365 changes what a state AI law currently requires of your business, and the litigation trying to change that could take years to resolve either way.
Here's the trap: a compliance program set to "wait and see" reads as prudent right up until an examiner, plaintiff's attorney, or state AG shows up asking why a law that was never actually struck down was never actually followed. Executive orders make headlines faster than they make case law, and the gap between those two speeds is exactly where the risk sits for anyone tracking US AI regulation right now.
What Executive Order 14365 actually does
President Trump signed Executive Order 14365 on December 11, 2025. Its stated purpose is to sustain "the United States' global AI dominance through a minimally burdensome national policy framework," and it pursues that through three levers rather than one, according to a client alert from Latham & Watkins.
The first and most consequential lever is Section 3: it directs the Attorney General to establish an AI Litigation Task Force whose job is to identify state AI laws and challenge them in court on the grounds that they unconstitutionally burden interstate commerce, are preempted by existing federal regulation, or are otherwise unlawful. The Department of Justice, led by Attorney General Pam Bondi, formalized that task force through an internal memorandum on January 9, 2026, with the group beginning work the following day.
The second lever is funding pressure. The order directs the Secretary of Commerce to evaluate whether states with what it calls "onerous" AI laws should lose access to federal Broadband Equity, Access, and Deployment (BEAD) funding, a $42.45 billion program overall that includes close to $1.86 billion earmarked for California alone.
The third lever is administrative: the order pushes the FCC to consider a federal AI disclosure standard and asks the FTC to issue a policy statement on how existing consumer-protection law applies to AI systems, both of which are slower-moving and have not produced final action as of this writing.
Section 1 of the order argues that state-by-state AI regulation "impermissibly regulate[s] beyond State borders, impinging on interstate commerce": the dormant commerce clause theory the task force is built to test. Colorado's AI Act is named directly in the order's text as an example of the kind of law the administration considers a target, specifically criticizing provisions the order says would "require entities to embed ideological bias within models."
What the order deliberately leaves alone
Section 8(b) of the order carves out three categories of state law from the preemption push entirely:
- Child safety protections: state laws aimed at protecting minors from AI-driven harms are explicitly out of scope.
- AI compute and data center infrastructure: state authority over the physical build-out of AI infrastructure, aside from generally applicable permitting reform, is preserved.
- State government procurement and use of AI: how a state buys and deploys AI for its own operations is left to the state.
If your compliance program touches any of those three areas, the executive order was never going to change your obligations there, regardless of how the litigation elsewhere turns out.
The litigation moved faster than "wait and see" anticipated
The most important thing to understand about the AI Litigation Task Force is that its first real courtroom involvement did not arrive as a DOJ-initiated lawsuit. It arrived as an intervention in a private company's case.
On April 9, 2026, Elon Musk's xAI filed suit against the state of Colorado in the U.S. District Court for the District of Colorado, challenging the Colorado AI Act (Senate Bill 24-205) on four constitutional grounds: that it compels speech in violation of the First Amendment, that it violates the dormant commerce clause by regulating conduct occurring entirely outside Colorado, that its carve-outs for diversity-promoting algorithms violate equal protection, and that the term "algorithmic discrimination" is unconstitutionally vague. xAI, organized under Nevada law and headquartered in Palo Alto, argued the law reached its model-development activity even though the company has no Colorado offices.
Two weeks later, on April 24, 2026, the Department of Justice moved to intervene in that case, marking the first time the federal government acted on the ground to challenge a state AI law under Executive Order 14365. On April 27, the court granted a joint motion staying enforcement of the Colorado AI Act, just over two months before its scheduled June 30 effective date.
That is a materially faster and more advanced timeline than an executive order signed four months earlier might have suggested. It is also, notably, not the pattern of "DOJ sues state, case grinds through years of litigation" that a headline about a "Litigation Task Force" implies. The actual mechanism was a private plaintiff's suit that the federal government joined.
The stay is not a ruling, and Colorado didn't wait around
Norton Rose Fulbright's analysis of the case is direct about how much the stay actually resolved: "Because the Court has not expressed any view on the merits of the case, that ultimate outcome is difficult to predict at this time." The disposition may turn on a genuinely novel constitutional question: whether an AI system's design choices count as protected expressive speech or regulable commercial conduct. Cato Institute's Jennifer Huddleston has described the case as AI's potential "Stratton-Oakmont moment," a reference to the early internet-era ruling that shaped decades of subsequent law, precisely because neither outcome is obvious yet.
Colorado did not sit still during the stay. Governor Jared Polis's AI Policy Working Group had already spent roughly six months in stakeholder consultation before proposing a narrower replacement framework in March 2026. On May 14, 2026, Polis signed SB 26-189, which strips out the original law's duty-of-care and formal risk-management requirements and replaces them with a disclosure- and transparency-focused framework, plus a 60-day pre-enforcement cure period. The new law takes effect January 1, 2027, and sunsets in 2030. Attorneys at Baker & McKenzie, writing in The Employer Report, called the stay "meaningful, but temporary, relief" and advised employers that "the most prudent course is a measured one": keep advancing compliance work, especially where it overlaps with other states' requirements, rather than standing down.
As of TechPolicy.Press's July 6, 2026 snapshot of the state AI landscape, states had enacted 109 AI-specific laws and 28 data center laws in 2026 to date. That is a slower pace than 2025, but it is not a freeze, and it is not evidence that state AI regulation is retreating in the face of the executive order.
Congress isn't resolving this either
If the executive branch's litigation strategy feels like an end-run, that's because it followed two failed attempts to get the same outcome through legislation.
In July 2025, Senate Republicans tried to attach a ten-year moratorium on state AI regulation to the budget reconciliation bill (the One Big Beautiful Bill Act). It didn't survive contact with the floor: the Senate voted 99-1 to strip the moratorium out, with only Senator Thom Tillis (R-NC) voting to keep it. The coalition against it ran from 17 Republican governors and 40 state attorneys general to the Heritage Foundation on one side and the Center for American Progress on the other, a rare alignment that signaled how little appetite existed for wiping out state AI law by federal fiat.
A second attempt followed in late 2025, this time trying to attach AI preemption language to the FY2026 National Defense Authorization Act. House Majority Leader Steve Scalise confirmed that provision would not make it into the final bill, after facing the same bipartisan resistance (state lawmakers, parents' groups, faith leaders, and unions among them) that sank the reconciliation-bill version months earlier. That failure is the direct backdrop to Executive Order 14365: a draft version of the order had circulated as early as November 2025, and once the NDAA route closed, the White House moved to litigation and funding leverage instead.
Two failed legislative attempts in under a year is meaningful context for how much weight the executive order alone should carry. An order can direct agencies to sue and can threaten funding, but it cannot rewrite a state statute, and Congress, so far, hasn't been willing to either.
What "wait and see" gets wrong
Paul Hastings put the operative legal reality plainly in a December 2025 client alert: "companies should continue to comply with applicable state AI laws because the Executive Order itself does not, and cannot, overturn existing state law: that can only be done by an act of Congress or the courts." White & Case's analysis reached a similar conclusion, noting that state AI laws "will likely not be impacted in the short term" by the order, since no federal AI law has been passed that would preempt them.
That's the piece a "wait and see" posture tends to miss: a law doesn't stop being enforceable because the federal government announced it plans to challenge it. Colorado's law is stayed by a specific court order tied to a specific case, not by the executive order itself, and even that stay is temporary and case-specific. Every other state AI law on the books, from Illinois's and Texas's AI employment-decision rules to California's ADMT regulations and Transparency in Frontier AI Act, remains fully enforceable right now, unless and until a court actually rules against it in a case that reaches that particular law.
And constitutional litigation of this kind rarely resolves in months. Even a case that loses decisively at the district court level typically has an appeal ahead of it, and Paul Hastings's own assessment anticipates a case that could reach the Supreme Court before anyone gets a durable answer to whether the dormant commerce clause theory holds up. Gibson Dunn's attorneys go further in their own analysis, arguing that a straightforward dormant commerce clause challenge is likely to fail on the merits because states retain broad authority to regulate conduct with in-state effects, and advising companies to keep "operat[ing] under the expectation that states will legislate" rather than betting on preemption. Firms are explicitly telling clients to prepare for a multi-year runway rather than a quick resolution in either direction. A compliance program built on the assumption that the litigation will resolve fast, in either direction, is planning against a timeline none of the lawyers actually litigating this expect.
A practical checklist for AI governance programs
None of this argues for treating the executive order as irrelevant. It argues for treating it as one input, not the input, into an AI governance program that keeps moving regardless of how the litigation lands.
- Don't pause work tied to a currently enforceable state law. If a state AI law applies to your business today and hasn't been stayed by an actual court order in an actual case naming your situation, it is still the law. Build to it.
- Track stays and rulings by state and case, not by headline. The Colorado stay applies to Colorado's AI Act, in that case, for now. It says nothing about California, Illinois, Texas, or Colorado's own replacement law once SB 26-189 takes effect January 1, 2027.
- Watch the three carve-outs separately. Child-safety AI rules, infrastructure/compute siting rules, and state procurement rules are outside the executive order's preemption push entirely. Don't let uncertainty elsewhere bleed into program areas that were never in question.
- Keep documentation current regardless of jurisdictional churn. Risk classifications, impact assessments, and system inventories built to a rigorous standard hold up whether a given state law survives, gets replaced (as Colorado's just was), or gets struck down years from now.
- Reassess on rulings, not on news cycles. A DOJ memo, a new client alert, or a state legislative session adjourning are all worth reading, but the thing that should actually move your compliance posture is a court ruling on the merits in a case that covers your situation.
That kind of tracking is hard to do by memory across a dozen states and one live constitutional fight. Secure Privacy's Privacy & AI Governance Platform includes an AI Governance module that registers each AI system your organization uses with its risk tier and maps it to the specific regulations that currently apply, so a Colorado stay or a Utah amendment shows up as a change to one system's applicable-law list rather than a reason to re-audit everything. The platform's Governance Maturity module also generates board-ready reports that can show, in one view, exactly which obligations are live, which are paused by a specific court order, and which never applied in the first place, useful the next time someone on the leadership team asks whether the executive order means the AI compliance budget can wait.
How this fits with the rest of your AI compliance calendar
This isn't happening in isolation. The EU AI Act's next major enforcement milestone lands August 2, 2026, and several US states have deadlines of their own running in parallel with the federal preemption fight; our California AI regulations guide walks through the CPRA's ADMT rules and the Transparency in Frontier AI Act obligations that are unaffected by anything discussed here. Organizations trying to run one coherent AI governance program across the EU AI Act, NIST's AI RMF, and ISO 42001 face a similar lesson: the frameworks don't wait for a single jurisdiction's litigation to resolve before their own clocks keep running, a point our EU AI Act vs. NIST AI RMF vs. ISO 42001 comparison covers in more depth.
The uncomfortable but accurate summary is that Executive Order 14365 has real teeth and a real, moving court case behind it, and none of that yet means what a "wait and see" compliance team might want it to mean. State AI law is enforceable until a specific ruling says otherwise, that ruling could be years away even after the current case resolves, and the two chambers of Congress that could settle this faster have already tried and failed twice.
FAQ
What is Executive Order 14365?
Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," was signed by President Trump on December 11, 2025. It directs the Department of Justice to create an AI Litigation Task Force to challenge state AI laws on constitutional grounds, and directs federal agencies to explore funding restrictions and administrative preemption tools, while explicitly leaving state child-safety, AI infrastructure, and government-procurement laws untouched.
Has any state AI law actually been struck down?
No. As of this article's publication, Colorado's AI Act (SB 24-205) has been temporarily stayed by a federal court as part of ongoing litigation (xAI v. Colorado, with DOJ intervening), and Colorado has since replaced it with a narrower law, SB 26-189, effective January 1, 2027. No court has issued a final ruling invalidating any state AI law under the theories the executive order advances.
Does the AI Litigation Task Force only target Colorado?
Colorado's AI Act was named in the executive order's text as an example and became the first case the task force intervened in, but the order directs the task force to identify and challenge any state AI law it considers inconsistent with federal policy, not Colorado's alone. Businesses operating under other state AI laws, including in California, Illinois, and Texas, should not assume they're outside the task force's scope simply because litigation hasn't reached those laws yet.
Should my company stop complying with state AI laws because of the executive order?
No. Multiple law firms tracking this litigation, including Paul Hastings and White & Case, advise clients to continue complying with applicable state AI laws, since an executive order cannot itself overturn state statutes. Only a final court ruling or new legislation can do that, and neither has happened for any state AI law as of this writing.
Why did Congress not just pass a federal AI preemption law instead?
Congress tried twice. A ten-year state AI moratorium was stripped from the 2025 budget reconciliation bill by a 99-1 Senate vote in July 2025, and a second attempt to add AI preemption language to the FY2026 National Defense Authorization Act was dropped before passage after facing similar bipartisan opposition. The executive order followed both failures.
How long will the Colorado AI Act litigation take to resolve?
There's no fixed timeline, and legal analysts tracking the case, including at Paul Hastings, expect it could take years and potentially reach the Supreme Court given the novel constitutional questions involved, particularly whether AI model design choices qualify as protected speech. The current stay is a preliminary procedural step, not a ruling on the merits.
Trying to plan your AI governance roadmap around when this litigation might resolve is a bet against a timeline the lawyers arguing the case won't commit to either. Book a demo of Secure Privacy's Privacy & AI Governance Platform to see how the AI Governance and Governance Maturity modules keep your program moving on the laws that are actually in force today, regardless of which ones are still being litigated.
