The document landed quietly on March 20. The White House’s National Policy Framework for Artificial Intelligence isn’t a regulation, an executive order, or an enforcement action. It’s a set of recommendations to Congress, seven of them, and its practical effect on any organization’s compliance obligations today is precisely zero.
That framing isn’t dismissive. It’s important. Because the gap between what this framework recommends and what currently governs AI deployment in the United States is where compliance teams are actually operating right now. Understanding that gap requires mapping who wants what, where they disagree, and what that means for planning decisions that can’t wait for legislative resolution.
What the Framework Actually Contains
Seven pillars, according to legal analyses of the document. The Morrison Foerster analysis confirms them directly: child protection; AI infrastructure and small business support; intellectual property; free speech; innovation; workforce preparation; and a federal approach to AI regulation that includes preemption of certain state laws. The whitehouse.gov document’s own language references preventing “a fragmented patchwork” of state regulation – the clearest signal of where the administration’s preemption intent sits.
The philosophy threading through the framework is what legal analysts characterize as “minimally burdensome”, a preference for federal standards that don’t layer excessive compliance costs onto AI development, particularly for smaller companies. AI infrastructure and small business support appearing as a named pillar is consistent with that orientation. So is the framing around innovation as its own policy area.
What the framework does not contain, as of early April: binding requirements, enforcement mechanisms, agency rulemaking timelines, or a Congressional co-sponsor. The JDSupra / Braumiller Law Group analysis frames it explicitly as a legislative agenda – a list of areas Congress is being asked to act on, not a record of action already taken. Legal analysts also note that a Commerce Department evaluation on state preemption, due by March 11, 2026, had not been publicly released as of early April. That evaluation was presumably meant to inform the preemption debate. Its absence is a gap in the legislative record.
The Preemption Pillar: What It Would Mean and What It Wouldn’t
Federal preemption of state AI law is the framework’s most consequential proposal, and the most contested one. According to legal analyses, the White House recommends preempting state AI laws that impose undue burdens on interstate commerce, while preserving state authority in specific carve-out areas including child protection and fraud prevention.
In practical terms, federal preemption would mean that a single federal AI compliance standard supersedes the patchwork of state requirements that currently applies to companies operating across multiple jurisdictions. That is genuinely attractive to legal and compliance teams managing multi-state programs. One standard is cheaper, faster, and less operationally fragmented than fifty.
The complexity is in the carve-outs. “Undue burden” is a legal standard, not a bright line. What constitutes an undue burden on interstate commerce is exactly the kind of question that gets litigated for years after legislation passes. The child protection and fraud prevention carve-outs mean that even under a preemption regime, organizations can’t simply discard state law analysis – they need to determine whether a given state requirement falls within a preserved category or within the preempted zone. That determination requires legal judgment, not just compliance checklists.
The Alvarez & Marsal analysis addresses this tension directly in its examination of the unsettled future of state AI laws under the framework.
The Competing Federal Visions
The White House framework isn’t the only federal AI proposal in play. Reports indicate that Senator Marsha Blackburn has introduced the “TRUMP AMERICA AI Act,” which legal analysts describe as a more prescriptive approach to federal AI regulation. Where the White House framework articulates principles and directs Congress to act, the Blackburn bill represents a more detailed legislative structure.
The two proposals share a federal approach, both assume that federal law, not state law, should govern AI, but they differ on how prescriptive that federal law should be. The “minimally burdensome” philosophy of the White House framework and a more detailed statutory structure represent real tension, not just rhetorical difference. Whichever approach advances in Congress will determine how granular compliance obligations actually become.
Then there’s the third position: Republican state lawmakers who have pushed back against federal preemption entirely. Reports indicate that some have urged the White House not to override state AI legislation. This isn’t a fringe position. States have moved faster than Congress on AI regulation. New York and Washington have enacted AI laws. Texas, Colorado, and others have active legislative processes. State lawmakers who built those frameworks have little interest in seeing them preempted by federal legislation that may be less specific, less enforceable, or less attuned to local conditions.
This is the political dynamic that makes clean federal preemption unlikely in the near term. The White House can recommend it. Congress has to pass it. And passing it requires support from legislators whose states have already acted, legislators who have political reasons to defend what their states built.
What Compliance Teams Must Do While Congress Decides
The Transparency Coalition’s April 3 legislative update and the ongoing legal commentary from firms tracking this space share a consistent message: state law is the operative compliance environment, and it will remain so until Congress acts and survives legal challenge.
That means three concrete things for compliance planning right now.
First: treat the White House framework as a signal about regulatory direction, not a change in current obligations. Use the seven pillars to anticipate where federal requirements may eventually land, particularly on intellectual property, child protection, and workforce, and stress-test your current state-by-state compliance posture against those anticipated areas.
Second: don’t assume preemption resolves the multi-state compliance problem anytime soon. Even if federal legislation passes, implementation timelines, agency rulemaking, and legal challenges mean multi-state compliance remains the reality for the foreseeable planning horizon. Organizations that have deferred state law compliance pending “clarity from Washington” are carrying real regulatory risk right now.
Third: track the Commerce Department evaluation. If and when it surfaces, it will be the most substantive federal input on the preemption question since the framework itself. It may reshape the legislative debate, or confirm what legal analysts already expect.
The White House framework represents the clearest articulation yet of where federal AI regulation is headed. But “headed” and “arrived” are separated by a legislative process that has no clear timeline and three competing visions fighting for the outcome. The compliance team that plans for where federal law might be in two years, while executing against where state law is today, is the one positioned to handle whatever Congress eventually produces.