Trump's National AI Framework: Federal Preemption, Deregulation, and the End of State AI Laws
Key Takeaways for Practitioners
On March 20, 2026, the Trump administration unveiled a proposed National AI Legislative Framework designed to eliminate the 'discordant mosaic' of state AI laws.
The Framework invokes federal preemption to nullify state AI regulations — potentially invalidating laws in California, Colorado, Texas, Illinois, and 40+ other states.
In January 2025, Trump revoked Biden's Executive Order 14110, eliminating AI safety protocols and red-teaming requirements for high-risk models.
The deregulatory bet: AI is too vital to national security and economic competitiveness to be constrained by state-level regulation.
Critics: the Framework clears the path for Big Tech without establishing federal protections to replace the state safeguards it eliminates.
Direct conflict with state tracker: 194 AI laws enacted across 45 states in 2025 alone — many directly in the Framework's preemption crosshairs.
The End of the Mosaic
On March 20, 2026, the Trump administration unveiled its official proposal for a unified national framework on artificial intelligence — a document that, if enacted, would represent the most significant restructuring of US AI governance since the technology emerged as a major economic and security concern.
The Proposed National AI Legislative Framework is not a technical guideline. It is a preemption strategy. Its central purpose is to eliminate what the administration calls the "policy barriers" created by fifty inconsistent state AI laws, replacing them with a single federal standard optimized for American competitive dominance over China.
The implications for practitioners advising AI companies, technology clients, and regulated industries are immediate and significant.
The Regulatory Landscape the Framework Seeks to Displace
What States Have Built
Our 50-State AI Tracker documents the scope of what the Framework is targeting: 194 AI laws enacted across 45 states in 2025 alone. These laws cover:
- Algorithmic discrimination in employment (Illinois, New York, Colorado)
- AI use in criminal justice — risk assessments, facial recognition (Virginia, Colorado, California)
- Health insurance AI prior authorization (Arizona, Indiana, Maryland, Nebraska, Texas)
- Deepfake regulations — electoral, sexual, commercial (30+ states)
- Government AI transparency and human oversight requirements (Kentucky, Montana, Utah)
- AI in education restrictions (Nevada, Illinois)
This is not a peripheral regulatory environment. It is the primary legal framework that most US AI deployments operate under today.
The Biden Baseline
The Framework explicitly displaces not just state law but also Biden-era federal policy. In January 2025, President Trump revoked Executive Order 14110 — Biden's October 2023 AI Executive Order that had established:
- Safety testing requirements for frontier AI models
- Red-teaming mandates for high-risk systems
- Transparency obligations for AI-generated content
- NIST AI Risk Management Framework as a federal reference standard
The new mandate: remove regulatory obstacles to American AI innovation.
The Preemption Theory
How Federal Preemption Works
The Supremacy Clause of the US Constitution establishes that federal law supersedes conflicting state law. Federal preemption operates through three mechanisms:
Express preemption: Congress explicitly states that federal law preempts state regulation in a field.
Field preemption: Federal regulation of a field is so comprehensive that it implicitly occupies the entire regulatory space, leaving no room for state law.
Conflict preemption: State law conflicts with federal law — either because compliance with both is impossible, or because state law frustrates the purposes of federal law.
The National AI Framework, as proposed, primarily relies on field preemption: the argument that AI development and deployment is an intrinsically interstate phenomenon with foreign policy dimensions that requires uniform national governance, leaving no space for state regulation.
The Constitutional Vulnerability
Field preemption requires congressional action — a President cannot unilaterally preempt state law through executive order or administrative guidance. The Framework is a legislative proposal, not an enacted statute. Its preemptive effect depends on Congress passing the proposed legislation.
This creates a significant gap between the administration's stated objectives and its current legal authority. Until Congress acts, state AI laws remain valid and enforceable. The Framework's preemptive force is prospective, not immediate.
ISSUE: Is Federal Preemption of State AI Law Constitutional and Wise?
The Case for Preemption
AI systems operate across state lines by design. A model trained in California, deployed on servers in Virginia, used by customers in 50 states, and governed by 50 different regulatory regimes creates genuine compliance dysfunction. The compliance cost of navigating 50 AI regulatory frameworks — with varying definitions of "high-risk AI," different disclosure requirements, and inconsistent enforcement regimes — falls disproportionately on smaller AI companies that cannot afford multi-state compliance infrastructure.
The national security argument is also real: if AI development in the US is constrained by regulatory fragmentation while China operates under a unified national strategy, the competitive disadvantage is measurable.
The Case Against
The states are not merely adding friction — they are filling gaps. The Framework preempts state AI laws but does not replace them with equivalent federal protections. California's ban on certain facial recognition uses, Colorado's algorithmic accountability requirements, and Virginia's criminal justice AI oversight rules would be invalidated without federal analogs.
This is the "race to the bottom" concern: federal preemption that eliminates state protections without substituting federal ones effectively deregulates the field entirely, leaving citizens with no recourse against AI harms that neither state nor federal law addresses.
What the National AI Framework Means for AI Practitioners Now
State laws remain valid and enforceable until Congress acts. The Framework is a proposal. Do not advise clients to cease state AI compliance on the basis of this proposal.
Monitor the legislative track. The Framework requires congressional enactment to have preemptive force. Track the proposed legislation's progress — and anticipate significant opposition from state attorneys general and civil rights organizations.
The 50-State Tracker remains essential. 194 enacted state AI laws govern current AI deployments. Until federal legislation is enacted and its preemptive scope is judicially defined, the state law landscape is the operative compliance framework.
Dual compliance is the prudent strategy. Design AI compliance programs that satisfy the most stringent applicable state requirements. If federal preemption ultimately eliminates those requirements, compliance excess costs nothing. If preemption fails or is narrowly construed, you're protected.
Document the regulatory conflict in client advice. The tension between the proposed federal framework and existing state laws creates genuine legal uncertainty. Client advice memos on AI compliance should address this uncertainty explicitly.
This analysis is based on publicly available government documents. It does not constitute legal advice. The regulatory landscape described is subject to rapid change.