On March 20, 2026, the White House released a National Policy Framework for Artificial Intelligence, a set of legislative recommendations urging Congress to preempt state AI laws it considers unduly burdensome and to replace a fragmented state landscape with a single national standard.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source The framework creates no legal obligations; it is a legislative blueprint, not enacted law, and companies should not treat it as an immediate compliance trigger.2 Holland & Knight 2026-03-24 The framework is non-binding legislative guidance, preempts state laws imposing undue burdens, and preserves state powers over children, fraud, and generally applicable law. Open source 3 Morgan Lewis 2026-03-24 Describes the framework as a blueprint not enacted law and names Senator Blackburn's discussion draft as the clearest legislative vehicle, with uncertain support. Open source Why it matters: it converts the posture of a December 2025 executive order into a concrete ask of Congress, and moves the preemption question into a body that has already failed once to freeze state rules. We assess, with moderate confidence, that the framework will set the terms of the 2026 preemption debate but will not by itself deliver federal preemption, because the Senate votes are not visibly present.3 Morgan Lewis 2026-03-24 Describes the framework as a blueprint not enacted law and names Senator Blackburn's discussion draft as the clearest legislative vehicle, with uncertain support. Open source

The drivers

The framework organizes its recommendations around six objectives: child safety and parental controls, community and ratepayer protection tied to data center electricity costs, intellectual property including a stance that training on copyrighted material does not itself violate copyright, free speech protections against government coercion of AI providers, innovation through regulatory sandboxes and no new standalone federal AI regulator, and workforce readiness.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source Beneath the list sits one governing idea: a uniform federal standard should displace state law wherever state rules impose what the administration calls undue burdens.2 Holland & Knight 2026-03-24 The framework is non-binding legislative guidance, preempts state laws imposing undue burdens, and preserves state powers over children, fraud, and generally applicable law. Open source

This did not arrive in a vacuum. The framework is the next step after Executive Order 14365, signed December 11, 2025, which directed the Commerce Department to identify onerous state AI laws within 90 days and established a Department of Justice AI Litigation Task Force, announced January 9, 2026, to challenge state measures the administration views as unconstitutional.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source The framework translates that executive posture into a request for statute. It preserves, on its face, state police powers: laws protecting children, laws against fraud, generally applicable laws, procurement decisions, and a state government's own use of AI all survive.2 Holland & Knight 2026-03-24 The framework is non-binding legislative guidance, preempts state laws imposing undue burdens, and preserves state powers over children, fraud, and generally applicable law. Open source What it targets is the layer of state law that reaches into how models are built and what liability developers carry for third party misuse.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source

Second order effects and the ledger

The mechanism that matters is not the framework itself but what it invites: a preemption statute. On the gains side, the largest frontier developers benefit most directly, because a single federal ceiling would relieve them of tracking and complying with a growing set of divergent state regimes, the compliance cost of which scales with the number of jurisdictions rather than the size of the model.2 Holland & Knight 2026-03-24 The framework is non-binding legislative guidance, preempts state laws imposing undue burdens, and preserves state powers over children, fraud, and generally applicable law. Open source The administration gains a coherent talking point that frames state regulation as the obstacle to competitiveness.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source

On the losing side sit state legislatures and attorneys general who have spent recent sessions building AI statutes, most visibly California, whose training data transparency and content provenance requirements are the archetype of the rules a preemption bill would displace.3 Morgan Lewis 2026-03-24 Describes the framework as a blueprint not enacted law and names Senator Blackburn's discussion draft as the clearest legislative vehicle, with uncertain support. Open source Consumer and safety advocates lose the state floor they have relied on while federal law stalled. The framework tries to soften this by carving out child protection and fraud, but the carve outs are narrow: they preserve enforcement of general harms while removing the developmental oversight that safety groups argue is where risk is actually set.2 Holland & Knight 2026-03-24 The framework is non-binding legislative guidance, preempts state laws imposing undue burdens, and preserves state powers over children, fraud, and generally applicable law. Open source

There is a subtler effect. By declining to create a new federal regulator and leaning on existing agencies, the framework signals that federal preemption here means less regulation, not a swap of state rules for equivalent federal ones.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source That reframes the debate: the question for Congress is not which government writes the AI rulebook, but whether a national rulebook that is deliberately light replaces state rules that are, in several states, heavier.3 Morgan Lewis 2026-03-24 Describes the framework as a blueprint not enacted law and names Senator Blackburn's discussion draft as the clearest legislative vehicle, with uncertain support. Open source

The counter-case

The main assessment could be wrong in two ways. First, the framework is non-binding, so its influence depends entirely on Congress acting, and the clearest existing vehicle, a discussion draft from Senator Blackburn, has uncertain Democratic support that will likely track election outlooks rather than policy merits.3 Morgan Lewis 2026-03-24 Describes the framework as a blueprint not enacted law and names Senator Blackburn's discussion draft as the clearest legislative vehicle, with uncertain support. Open source If that support fails to materialize, the framework becomes a statement of intent with no statute behind it, and state law continues to accrete. Second, the opposite risk: the AI Litigation Task Force gives the executive branch a route to weaken state laws through the courts without any congressional vote at all, which would make the framework's legislative ask less central than it appears.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source For the thesis that this framework sets but does not settle the debate to fail, either Congress would have to move faster than its recent record suggests, or litigation would have to do the preemption work first.

What to watch

  • A preemption bill gets a committee vote. If a federal preemption measure built on the framework reaches a House or Senate committee markup before the end of 2026, the legislative path is live; no markup by year end signals the framework stalled as guidance.3 Morgan Lewis 2026-03-24 Describes the framework as a blueprint not enacted law and names Senator Blackburn's discussion draft as the clearest legislative vehicle, with uncertain support. Open source
  • The Commerce state law inventory surfaces. EO 14365 required Commerce to identify onerous state laws; publication of that list would name the specific state statutes in the administration's sights and sharpen the fight.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source
  • The AI Litigation Task Force files. A first DOJ suit challenging a state AI law within the next two quarters would show the executive branch pursuing preemption through courts in parallel with the legislative ask.1 WilmerHale 2026-03-23 The March 20 2026 framework, its six objectives, the preemption ask, and the link to EO 14365 and the DOJ AI Litigation Task Force announced January 9 2026. Open source
  • A large state enacts anyway. If California or a comparable state passes a significant new AI statute in its 2026 session despite the framework, the state patchwork keeps growing and the pressure the framework describes intensifies rather than resolves.2 Holland & Knight 2026-03-24 The framework is non-binding legislative guidance, preempts state laws imposing undue burdens, and preserves state powers over children, fraud, and generally applicable law. Open source