The White House Wants to Own AI Regulation. Thirty-One States Are Refusing.

The Trump administration's push to establish federal primacy over AI governance has run into a coordinated wall of state-level resistance, setting up what legal scholars say could become the defining constitutional battle of the AI era. The federal government wants to set the rules. Most states are writing their own anyway.

The conflict escalated after the White House issued an AI executive order on June 2 calling on AI developers to voluntarily submit their models for federal review before deployment — framing this as a national security and competitiveness measure. But bundled with that order is a broader White House argument: that AI regulation should be a federal function, and that state-level AI laws create a fragmented compliance environment harmful to American innovation.

States aren't buying it.

How Many States and What They're Doing

At least 31 states are advancing AI legislation in their current sessions, according to tracking by Vorys Law. The categories span consumer protection, hiring discrimination, healthcare AI use, facial recognition, and algorithmic accountability. This is not a wave of identical copycat bills — these are jurisdictions with different priorities, different economies, and different political calculi taking independent action.

The most consequential state law now in effect: Colorado's AI Act, which takes effect June 30, 2026. Colorado's law is the broadest state AI regulation yet passed in the U.S., covering high-risk AI systems in employment, education, credit, healthcare, housing, and insurance. It places affirmative obligations on both AI developers and deployers — not just the companies building models, but the businesses using them.

Colorado isn't alone in the near-term window. Several other states are approaching final passage on similar frameworks this session.

The Preemption Argument — and Why States Reject It

Federal preemption — when federal law supersedes state law — is a standard legal tool. But it requires either a specific congressional act with preemption language or a constitutional conflict with federal authority. The White House's current AI order is voluntary guidance, not legislation. It asks companies to submit models for review; it cannot legally compel them to do so, and it certainly cannot legally prohibit states from passing their own binding laws.

State attorneys general across multiple states have already signaled they view the White House framing as legally unsupported. The argument is straightforward: absent an act of Congress explicitly preempting state AI regulation, states retain their traditional police powers over consumer protection, employment law, and civil rights enforcement — all domains touched by current AI deployments.

Legal scholars say this is heading to court. The specific trigger is likely to be Colorado's June 30 effective date, which will produce immediate compliance conflicts for companies operating nationally. A company that deploys an AI hiring tool designed to meet federal guidance could still face state liability under Colorado law for failing to conduct the impact assessments Colorado requires.

According to Vorys Law's analysis, legal scholars believe this conflict could reach the Supreme Court within 18 months — and the outcome could define not just AI governance but the broader question of how federal and state authority is allocated in emerging technology regulation.

Why States Are Acting Independently

The states' posture isn't pure defiance. It reflects a practical reality: Congress has not passed a comprehensive federal AI law, and there is no immediate sign that it will. In that vacuum, states are responding to constituent pressure, industry lobbying (in both directions), and specific harms being documented in their jurisdictions.

Several state-level efforts are also explicitly better funded and more technically informed than federal counterparts. California's AI legislative staff have engaged directly with academic AI safety researchers. Colorado's law reflects significant input from the Algorithmic Justice League and civil rights organizations who argued that federal guidance had not adequately addressed discrimination risks in automated decision-making.

There is also a political dimension that crosscuts the usual partisan lines. Some Republican-led states are advancing AI laws focused on prohibiting government use of AI facial recognition. Some Democratic-led states are advancing laws focused on employer liability for AI hiring decisions. The issue doesn't align cleanly with federal party positioning, which reduces the likelihood of a quick congressional resolution.

What This Means for Businesses

For any company deploying AI in consumer-facing or employment contexts in the U.S., the current environment is the worst possible outcome: no single rule set to follow, a patchwork of state obligations, and federal guidance that lacks enforcement teeth but still creates compliance overhead.

The practical risk matrix:

• Colorado (June 30): High-risk AI deployers must conduct impact assessments, maintain human review processes, and provide notice to affected individuals. Developers must make documentation available to deployers.
• Other states in process: Requirements vary but the trend is toward impact assessments, algorithmic transparency, and anti-discrimination provisions.
• Federal level: Voluntary model submission with no current legal obligation and no clear enforcement mechanism.

Companies that had structured their AI compliance programs around the premise that federal guidance was imminent and would preempt state action need to revisit that assumption. Colorado is live in 25 days. Waiting for federal clarity is no longer a strategy.

What to Watch

The June 30 Colorado effective date is the immediate trigger point. Expect legal challenges from industry groups within weeks of that date, most likely framing the issue as an unconstitutional burden on interstate commerce. The second key milestone is whether Congress moves a federal AI bill before year-end — if it does, the preemption question moves to legislative drafting; if it doesn't, the courts become the venue by default.

By Hector Herrera | NexChron | June 5, 2026