The fight is over who decides

The United States has no comprehensive federal law governing artificial intelligence. In that vacuum, states began to legislate — Colorado, California, Utah, Texas, and others passed measures touching algorithmic discrimination, deepfakes, transparency, and high-risk automated decisions. The same pattern that produced a 50-state patchwork in data privacy is now repeating in AI.

The current fight is not, at bottom, about whether AI should be regulated. It is about who gets to regulate it — and the legal mechanism at the center of that fight is preemption: the doctrine that federal law can displace conflicting state law. Over the past year, the federal executive branch has moved aggressively to clear the field of state AI regulation, while Congress has repeatedly declined to do so by statute. The result is an unusually direct test of federalism.

Why preemption is the lever

Under the Constitution’s Supremacy Clause, valid federal law overrides conflicting state law. Congress can preempt state law expressly, by writing a preemption clause into a statute, or impliedly, where state law conflicts with federal law or intrudes on a field Congress meant to occupy. Courts, separately, can strike down state laws that unconstitutionally burden interstate commerce under the dormant Commerce Clause.

This matters because state laws are, for now, the only binding AI rules most companies face. If those laws are preempted or enjoined and no federal standard replaces them, the practical outcome is not a uniform national rule — it is no enforceable rule at all. That is the structural stake. Preemption can mean replacing a patchwork with a single standard, or it can mean deregulation by subtraction.

What Congress declined to do

The first attempt to preempt state AI law ran through Congress. A provision in the 2025 budget reconciliation bill (“One Big Beautiful Bill Act”) would have imposed a 10-year moratorium barring states from enforcing their own AI laws. The Senate stripped it out on July 1, 2025, by a 99-1 vote — a rare bipartisan rejection, with opposition from both progressives wary of deregulation and conservatives wary of overriding state authority.

House Republican leaders revived the idea in late November 2025, seeking to attach a state-AI-law moratorium to the FY2026 National Defense Authorization Act. That effort also failed: the compromise NDAA released in December contained no moratorium, and the bill was signed into law on December 18, 2025 without it. Twice, through the chamber best positioned to weigh state interests, Congress declined to preempt.

What the executive branch did instead

On December 11, 2025, the President signed an executive order titled “Ensuring a National Policy Framework for Artificial Intelligence.” Having failed to secure preemption by statute, the order pursues it through executive tools:

• A DOJ AI Litigation Task Force. The order directs the Attorney General to establish, within 30 days, a task force to challenge state AI laws the Attorney General deems to unconstitutionally regulate interstate commerce, to be preempted by federal regulation, or to be otherwise unlawful. Attorney General Bondi formally established it in January 2026.
• Conditioning federal funds. The Secretary of Commerce is directed to make states with disfavored AI laws ineligible for Broadband Equity, Access, and Deployment (BEAD) program funds — billions in broadband money from the 2021 infrastructure law. Other agencies are told to consider conditioning their discretionary grants similarly.
• Agency rulemaking as a preemption vehicle. The FCC is directed to consider a federal AI reporting and disclosure standard intended to preempt conflicting state law; the FTC is directed to address when state laws regulating AI outputs conflict with federal prohibitions on deception.
• A legislative recommendation. Senior officials are to draft a federal preemption bill, preserving narrow carveouts for child-safety protections, AI infrastructure, and government use of AI.

The task force moved quickly. In April 2026, the DOJ intervened in xAI v. Colorado officials, a suit challenging Colorado’s AI Act — the first time the federal government has intervened against a state AI law. A federal magistrate judge stayed enforcement of that law later in April.

It is worth being precise about what an executive order can and cannot do. An executive order is not itself law and cannot preempt a state statute on its own. Its force here is indirect: litigation that asks courts to find preemption, and financial pressure that asks states to back down. Whether the underlying preemption theories succeed is a question for the courts, and several legal commentators have questioned how far they reach.

What Colorado signals

Colorado was the first state to enact a comprehensive AI law: the Colorado AI Act (SB 24-205), signed in May 2024, requiring developers and deployers of high-risk systems to manage the risk of algorithmic discrimination through impact assessments, disclosures, and risk-management programs. Its effective date was repeatedly delayed.

On May 14, 2026, the governor signed SB 26-189, which repeals that 2024 law and replaces it with a narrower framework. The new law drops the duty of care, the algorithmic-impact-assessment mandate, and the risk-management-program requirement, substituting transparency and disclosure obligations: developers must document intended uses and training data, and deployers must disclose AI use and explain adverse decisions, with consumers able to request correction and human review. It takes effect January 1, 2027.

Colorado officials attribute the rewrite mainly to state-level concerns about the original law’s enforceability and compliance burden, following a gubernatorial workgroup’s recommendations. But the retreat of the nation’s first comprehensive state AI law — under federal litigation pressure and amid a broad federal preemption campaign — is the clearest signal yet of which direction the field is moving.

What is actually at stake

The honest framing is not “uniform national standard vs. messy patchwork.” It is this: federal preemption removes state authority now, while a federal replacement standard exists only as a proposal. Until Congress acts, preemption that succeeds in court leaves a regulatory vacuum rather than a national rule. The substantive questions state laws were trying to answer — algorithmic discrimination in hiring and lending, transparency about automated decisions, recourse for people harmed by AI systems — do not disappear because the laws addressing them are struck down.

What to ask your representatives

• Do they support a federal AI standard, and if so, would it be a floor that states can exceed, or a ceiling that preempts stronger state protections?
• Should federal funds like BEAD broadband money be conditioned on a state’s AI-policy choices?
• If state AI laws are preempted or enjoined, what enforceable protections against algorithmic discrimination and opaque automated decisions would remain?
• Do they believe AI preemption should be decided by Congress, by executive order, or by the courts?