The White House released its National Policy Framework for Artificial Intelligence on March 20, 2026, outlining what it wants a unified federal AI regulatory system to look like. The document is a set of legislative recommendations to Congress, not an executive order, not a rule, and not a law. That distinction matters for anyone tracking compliance obligations: nothing in this Framework changes what companies must do today.
What the Framework recommends is still significant. According to legal analysis from Holland & Knight, who reviewed the document, the Framework prioritizes “targeted federal preemption” and explicitly cautions against “vague standards, open-ended liability and fragmented state regulation.” It also addresses child safety, community protections, free speech, innovation, and workforce readiness as core policy priorities. The Framework reportedly includes dozens of specific legislative recommendations, according to Global Policy Watch, though that count has not been independently confirmed for this package.
The Framework is organized around several policy priorities, according to Holland & Knight and other legal analysts who reviewed the document. The precise structure, including how many pillars or thematic areas the document uses, has not been independently confirmed from primary text review. The full PDF of legislative recommendations is available at the White House website, and any characterization of its specific provisions should be verified against that document before being treated as settled.
This Framework doesn’t stand alone. A separate White House executive order from December 2025, titled “Ensuring a National Policy Framework for Artificial Intelligence”, preceded it and operates through a different legal mechanism entirely. The December EO is a directive to federal agencies. The March Framework is an appeal to Congress. They’re related but distinct instruments, and conflating them is a compliance error. Legal analysis of the December EO suggests it directs federal agencies to address state AI laws that conflict with national policy, though the specific operational language of that order requires full text review to characterize precisely.
The political context behind both documents is clear enough. States haven’t waited for federal action. Roughly 25 AI laws have been enacted across US states in 2026 alone, according to the legislative tracking platform Plural Policy, a figure that has accelerated sharply since mid-March. California issued its own AI executive order as recently as March 30. The Framework’s federal preemption language is a direct response to that legislative acceleration.
What it is. A White House blueprint recommending that Congress pass a unified federal AI law preempting state regulations, addressing child safety, free speech, innovation, and AI workforce readiness.
What it isn’t. A law. A rule. A binding requirement. Companies don’t have new compliance obligations because of this document.
What to watch. Whether Congress acts on any of the Framework’s recommendations, and at what pace. The Framework creates political momentum, but legislative calendars move slowly. Meanwhile, state laws keep passing. The gap between federal aspiration and legislative reality is where compliance professionals will spend the next twelve to eighteen months.
The Framework’s publication tells you where the White House wants to go. The states’ legislative output tells you where companies are being asked to comply right now. Those two vectors don’t yet point in the same direction.