Two regulatory frameworks. One compliance team. Pick one.
That’s the position AI compliance professionals at multi-state companies are in right now. The Trump administration’s National Policy Framework for Artificial Intelligence, released March 20, 2026, wants Congress to establish a single federal standard that preempts conflicting state AI laws. States with their own AI legislation, California, Colorado, and others with bills in progress, aren’t interested in standing down.
The framework itself isn’t law. It’s a four-page set of legislative recommendations. According to Morrison Foerster’s analysis, the document spans seven pillars: child protection, AI infrastructure and small business support, intellectual property, censorship and free speech, enabling innovation, workforce preparation, and preemption of state AI laws. Every provision requires Congressional action. Congress hasn’t taken that action. That gap, between what the administration wants and what currently exists as law, is where compliance teams are living right now.
The Preemption Conflict Map
The framework’s preemption proposal isn’t abstract. Specific state laws are in the crosshairs.
Colorado’s AI Act is already on the books, with compliance requirements for high-risk AI systems. California’s Governor signed an executive order on April 1, 2026, covered separately in today’s regulation briefing, that establishes independent AI oversight mechanisms for state government operations, including a requirement that California independently review federal supply-chain risk designations for AI businesses, according to Washington Post reporting. These aren’t hypothetical state assertions of authority. They’re active.
The administration’s framework proposes to preempt state laws that “impose undue burdens” while preserving state police powers. The line between a “burden” and a legitimate state regulatory interest is exactly the kind of question courts will eventually need to answer – and the administration appears to anticipate this. The December 2025 Executive Order that preceded the framework deployed an AI Litigation Task Force with a specific mandate to challenge state AI laws the administration considers inconsistent with federal policy. That task force is already operational.
The named stakeholders and their positions:
Trump Administration: Federal preemption is the stated goal. The seven-pillar framework represents the full legislative ask. The AI Litigation Task Force is the enforcement mechanism that operates while Congress deliberates.
California: The April 1 executive order signals Sacramento’s intent to maintain independent AI oversight regardless of federal direction. The independent supply-chain review provision is a direct counter to federal designation authority.
Colorado: Has existing AI Act compliance obligations for high-risk systems. Federal preemption, if enacted, would override these, a significant policy reversal for a state that moved early on AI governance.
Congress: Has not introduced legislation to codify the framework’s provisions. Without a bill, the framework is aspirational. The timeline for Congressional action is genuinely unknown.
Compliance teams at multi-state companies: Caught between two regulatory regimes, neither of which is fully in force. Must make resource allocation decisions now.
What Congress Would Actually Have to Do
Converting the administration’s framework into law isn’t a simple ratification. Each of the seven pillars represents a distinct legislative undertaking. The preemption provision alone would require defining what constitutes an “undue burden,” establishing which state laws are preempted and which aren’t, and likely surviving legal challenges from states asserting Tenth Amendment authority.
The administration acknowledged in the December 2025 Executive Order that legislative recommendations would be developed and submitted to Congress. That process produced the March 20 framework. The next step, actual legislation, requires sponsors, committee hearings, floor votes in both chambers, and reconciliation of differences between House and Senate versions. In a divided political environment, that’s a multi-year process at minimum, not a quarterly one.
Three Scenarios for Compliance Teams
Scenario 1, Federal preemption passes: Congress enacts legislation that establishes a uniform federal AI standard and supersedes conflicting state laws. Companies that built only to state-specific standards face rebuilding to federal requirements. Companies that built only to a federal standard are vindicated. Timeline: 18 months at minimum, more likely 3+ years.
Scenario 2, States hold their ground: Congress doesn’t act, or a preemption bill fails. State AI laws, Colorado, California, and any enacted during the delay, remain the operative compliance framework. Companies that waited for federal clarity are behind. Companies that built to state standards are ahead.
Scenario 3, Stalemate continues: The most likely near-term scenario. Congress debates but doesn’t pass comprehensive preemption legislation. The AI Litigation Task Force challenges specific state laws in court, creating a patchwork of rulings with no uniform standard. Compliance teams must track both federal enforcement signals and state-level developments simultaneously.
The framework’s copyright provision adds another layer. The administration states that training AI models on copyrighted material does not violate copyright law, according to analyses from Vital Law and Pinsent Masons, though courts are actively working through contrary arguments. This position, if adopted by Congress, would represent a significant shift in intellectual property policy with implications well beyond AI.
What to Watch
Three signals will indicate which direction preemption is actually heading.
First, watch for Congressional bill introductions. The framework exists. A bill to implement it doesn’t yet. The moment a preemption bill is introduced with genuine sponsor support, the compliance calculus changes.
Second, watch the AI Litigation Task Force. Its first legal challenges to state AI laws will reveal the administration’s enforcement priorities, which state laws it considers most problematic and where it’s willing to spend political capital.
Third, watch California. California’s response to federal preemption pressure has historically shaped national policy outcomes. If Sacramento enacts additional AI legislation that directly conflicts with the federal framework, the legal showdown accelerates.
TJS Synthesis
The administration’s framework is best understood as a declaration of intent, not a change in law. For compliance teams, the practical implication is this: the safest path is building to the stricter of the applicable standards, which, right now, means state law, while maintaining the architectural flexibility to adapt if federal requirements materialize.
That’s more expensive. It’s also the only position that doesn’t require betting on Congressional timing.