A framework isn’t a law. That sentence should anchor every internal conversation compliance teams have about the White House National Policy Framework for Artificial Intelligence, released on March 20, 2026.
The distinction matters because the framework’s most consequential recommendation, that Congress preempt state AI laws that “impose undue burdens”, has been widely covered as though it resolves the federal-state AI governance question. It doesn’t. It opens a legislative argument. And while that argument plays out, every state AI compliance obligation your organization currently holds remains fully enforceable.
What the Framework Is, and Isn’t
The framework was directed by a December 11, 2025 executive order and represents the deliverable from that direction: a formal statement of what the administration wants federal AI law to look like. It contains more than two dozen legislative recommendations organized across seven thematic policy areas. It is not itself an executive order. It creates no compliance obligations. It confers no safe harbors.
As Cooley’s March 25 client alert described it, this is “the most concrete statement yet of where the administration wants Congress to take federal AI policy.” That framing is precise and worth internalizing. It tells you the destination the administration is aiming for, not the road it has traveled.
A parallel signal arrived in the same week. Senator Marsha Blackburn (R-TN) announced the TRUMP AMERICA AI Act discussion draft, described as intended to “codify” the president’s preemption executive order. This draft has not been formally introduced as legislation. It is a discussion document, meaningful as a directional signal, not as an operative requirement.
The Preemption Mechanism: What Would and Wouldn’t Be Superseded
Federal preemption of state law is a well-established legal mechanism. When Congress acts, it can explicitly preempt state law (express preemption) or create a regulatory scheme so comprehensive that state law is implicitly displaced. The framework advocates for express preemption of state AI laws that place “undue burdens” on AI development and deployment.
The scope of that phrase is the critical variable. Legal analysts reviewing the framework report that carve-outs would include state authority over general child protection laws, consumer protection, fraud, data center zoning, and a state’s own use of AI. The full list of carve-outs, and what qualifies as an “undue burden,” would ultimately be defined by whatever legislation Congress passes. The framework signals the direction; Congress writes the boundary.
This uncertainty is not a reason to dismiss the preemption question. It is a reason to audit your current state AI compliance portfolio now, before the legislative debate sharpens. Organizations that have mapped their state law obligations are positioned to assess preemption impact quickly. Those that haven’t are starting from scratch if Congress moves fast.
The State Law Landscape: A Growing Target
The framework’s preemption recommendation lands in an environment where a growing number of states have enacted or are actively advancing AI-specific legislation. On the same day the framework’s implications were being reported, Washington State Governor Bob Ferguson signed two AI bills into law, a chatbot safety requirement and an AI-generated content disclosure requirement. That is not a coincidence of timing that should be treated as background noise. It is a precise illustration of the dynamic the framework is trying to resolve.
States aren’t waiting for federal clarity. The White House’s recommendation acknowledges this implicitly: you don’t need a preemption recommendation if states aren’t already acting. Compliance teams should not assume that the administration’s preference for federal uniformity will translate to legislative action on any particular timeline.
Stakeholder Positions: Who Wants What
The preemption debate involves constituencies with sharply different interests.
The administration has stated its position clearly: federal uniformity reduces compliance costs for AI developers, accelerates deployment, and prevents a state-by-state fragmentation that could impede U.S. competitiveness. The framework’s preemption recommendation reflects that view.
AI companies, particularly large developers operating across multiple states, generally favor federal preemption. A single national standard is simpler and cheaper to comply with than fifty different state regimes. This preference is not universal; some compliance structures built around state requirements would need to be dismantled under federal preemption, which carries its own transition cost.
State legislatures are unlikely to yield their authority without resistance. States have moved on AI legislation partly because federal action has been slow. Washington State’s March 26 signing is a demonstration of that posture. State legislators and attorneys general who have built AI oversight infrastructure have institutional and political incentives to defend it against federal preemption.
Compliance professionals face the sharpest near-term uncertainty. Federal preemption, if enacted, would require rebuilding state-specific compliance programs around a new federal standard. The timing, scope, and carve-outs of that standard are unknown. Building for preemption before it exists is premature; ignoring state obligations in anticipation of preemption is negligent.
What Compliance Teams Should Do Now
The practical posture is straightforward, even if the legal environment isn’t.
First, continue to honor all existing state AI compliance obligations. The framework creates no safe harbor, no transition period, and no preemption effect. Washington State’s new laws are enforceable from their effective date regardless of what Congress does.
Second, map your state AI law exposure. Identify which state AI laws currently apply to your organization’s AI systems, what they require, and when those requirements take effect. This mapping will serve two purposes: it gives you a current compliance baseline, and it becomes the input to your preemption impact analysis if federal legislation advances.
Third, monitor the Blackburn draft’s progress. A discussion draft that explicitly aims to codify the administration’s preemption executive order is the legislative vehicle most likely to carry preemption language forward in this session. Its advancement, from discussion draft to introduced bill to committee action, is the trigger that warrants a deeper compliance reassessment.
Fourth, do not build compliance programs around anticipated preemption. Companies that defer Washington State compliance because federal preemption “might happen” are accepting regulatory risk on a speculative basis. That is not a defensible position if an enforcement action arrives before federal legislation does.
What to Watch
The legislative timeline is the key variable. Watch for: formal introduction of the Blackburn draft; committee assignments and hearing schedules for any AI preemption legislation; statements from state attorneys general on the administration’s preemption intent; and whether additional states enact AI laws that could become preemption targets.
The copyright question is also moving in parallel. The framework’s position, that AI training on copyrighted material is lawful and fair use should be resolved by courts, connects to active litigation and to the Supreme Court’s recent declination to hear Thaler v. Perlmutter. The administration has taken a side. The courts haven’t finished.
The Insight That Matters
Federal preemption of state AI laws is a policy argument, not a legal outcome. The White House has made the argument clearly and formally. That is worth understanding. What it doesn’t do is change the compliance obligations that exist today, in Washington State or anywhere else a state AI law is on the books. The organizations best positioned to navigate whatever Congress does next are the ones that have already done the work on what states require now.