The Framework’s Preemption Mechanism: What It Actually Says
Federal preemption is a legal doctrine, not a policy preference. When federal law explicitly preempts state law, state requirements become unenforceable. The White House Framework’s preemption proposal isn’t that simple, and the complexity is where compliance planning gets difficult.
The Framework released March 20, 2026 calls for targeted federal preemption: specifically, preemption of state AI laws that “unduly burden” AI development. Not all state laws. Not a blanket federal override. State laws that survive the “undue burden” test stay intact. The Framework explicitly preserves states’ traditional police powers, child protection, fraud enforcement, and similar public safety functions, per Holland & Knight’s legal analysis of the document.
That carve-out is significant. It means a state law requiring AI systems to disclose when content is AI-generated might survive preemption, it’s hard to characterize consumer transparency as an undue burden on development. A state law requiring pre-launch safety assessments for general-purpose AI systems might not survive, it imposes a compliance process that could be characterized as burdening development. Where any given state law falls depends on how “undue burden” gets defined, and right now, no one has defined it.
The Framework is nonbinding. Congress must enact legislation for preemption to take legal effect. The K&L Gates analysis notes the Framework follows earlier executive action by President Trump on AI in December 2025. The specific content of that earlier action isn’t fully confirmed in available primary sources, but the trajectory is consistent: the Administration has been building toward a federal framework that consolidates AI regulation at the national level.
Who’s Pulling in Which Direction
Three distinct positions are shaping the federal AI landscape right now. They’re not fully aligned, even among nominal allies.
The White House Framework advocates for a unified federal approach with targeted state preemption, while preserving state police powers. Its priority areas, child safety, community protections, free speech, innovation, and workforce readiness, suggest a deliberate attempt to build bipartisan appeal. Intellectual property appears as a priority area, though the specific recommendations haven’t been fully detailed in available public summaries.
Senator Blackburn’s TRUMP AMERICA AI Act is the most relevant Congressional vehicle in the current conversation. Holland & Knight’s alert addresses both the Framework and the Blackburn bill in the same analysis, indicating a relationship between the two, but also noting apparent divergence on copyright. The specific nature of that divergence isn’t fully confirmed in available source material and must be characterized as such: the Administration and Senator Blackburn aren’t fully aligned on what federal AI legislation should say about intellectual property, according to Holland & Knight’s legal interpretation. That’s a meaningful fault line. IP protection is a priority area for technology companies, content creators, and the creative industries, and if the primary Senate vehicle diverges from the White House on this point, the legislative path gets more complicated.
State legislatures are the third force. California’s SB 53, signed by Governor Newsom on March 30, the same week the Framework was being analyzed across the legal and policy world, is the most direct evidence that states aren’t waiting for federal action. SB 53 requires large frontier AI developers to publicize safety frameworks and transparency reports, and mandates incident reporting for what the law defines as catastrophic risks. The specific thresholds are per reporting by Mustang News and should be confirmed against the official bill text at California Legislative Information. The point isn’t the specific numbers, it’s that the law exists, it’s signed, and it creates compliance obligations right now, without waiting for Congress.
California is not alone. It’s simply the most prolific. More than a dozen states had active AI-related legislation in some form as of early 2026. The Framework’s preemption push may accelerate state legislative activity in the near term, as state legislatures try to lock in requirements before a federal law could override them.
The Compliance Planning Problem
For legal and compliance teams at AI companies, this situation creates a genuine planning dilemma. The Framework presents two incompatible futures: one where Congress enacts federal AI legislation with preemption provisions, and one where Congress doesn’t act (or acts years from now), leaving a patchwork of state laws as the operative regulatory environment.
Building state-specific compliance programs is expensive. A company operating in ten states with active AI legislation needs to track, implement, and maintain ten distinct sets of requirements, requirements that may conflict with each other. The case for betting on federal preemption is that this investment becomes unnecessary if a federal law supersedes state requirements.
The case against betting on federal preemption is everything about how Congress works. Federal AI legislation has been attempted repeatedly without success. The Framework is nonbinding. The Senate vehicle has apparent divergence with the White House on at least one key issue. Bipartisan consensus on AI legislation, while theoretically possible given the Framework’s cross-cutting priority areas, hasn’t materialized in prior sessions.
The practical guidance that follows from this analysis: state laws are the operative legal reality until further notice. SB 53 is in effect. Companies meeting California’s threshold for “large frontier AI developers” have compliance obligations now. The White House Framework is a signal about where the Administration wants to go, it’s not a reason to defer state compliance work.
What “Congressional Action Required” Actually Means
The Framework is a recommendation document. Its preemption language only becomes law when Congress passes legislation incorporating it, and the President signs that legislation. This process involves committee referral, hearings, markup, floor votes in both chambers, and conference reconciliation if House and Senate versions differ. None of that has happened. A timeline for it happening hasn’t been established.
The bipartisan dynamics are genuinely uncertain. AI legislation has drawn interest from members of both parties, but for very different reasons, some focused on safety and accountability, others on innovation and competitiveness, others on copyright and creator protections. The Framework attempts to build a coalition by addressing multiple concerns, but the Blackburn bill’s apparent copyright divergence suggests that coalition isn’t fully formed even within the Republican caucus.
Characterizing a Congressional timeline with any specificity would be speculation. The honest assessment is: federal AI legislation may happen in this Congress, or it may not. The “undue burden” standard, if it ever becomes law, will almost certainly be litigated, its meaning won’t be settled by the legislation alone.
What to Watch
Five developments will shape whether the Framework moves from signal to law:
Congressional hearings on federal AI legislation. Committee activity, specifically the Senate Commerce Committee and House Energy and Commerce Committee, will indicate whether the Framework is getting traction or stalling. Watch for hearing schedules and witness lists.
The Blackburn bill’s trajectory. Whether it gains co-sponsors, how it’s amended, and whether the copyright divergence with the White House Framework gets resolved will signal how unified the legislative coalition is.
State legislative responses. If state legislatures interpret the preemption push as a reason to move faster, a wave of state AI laws before federal legislation passes would make preemption both more politically contentious and more practically necessary.
Legal challenges to SB 53. If California’s new law faces a preemption challenge in federal court, the litigation would force a judicial answer to questions the Framework leaves open, including what “undue burden” means in practice.
Definition of “undue burden.” This is the threshold question. Any proposed federal legislation that includes preemption language will have to define this standard. How it’s defined determines which state laws survive. Watch the legislative text carefully when it appears.
TJS Synthesis
The White House Framework and California’s SB 53 represent two legitimate responses to the same problem: AI systems are increasingly consequential, and the rules governing them are still being written. The Framework bets on federal consolidation as the right answer. California bets that states need to act now, before federal consolidation arrives.
Both can be right simultaneously, and that’s the compliance challenge. Companies subject to California law have legal obligations today. Companies planning for the medium term have to account for the possibility that those obligations get preempted, modified, or layered with additional federal requirements. The rational response is neither to ignore state law waiting for federal action, nor to build state-specific programs as if federal preemption is impossible.
Build for what’s law. Plan for what’s coming. The Framework tells you what the Administration wants. SB 53 tells you what’s required right now. Those are different documents with different legal weights, and treating them accordingly is the first step in a defensible compliance strategy.