White House AI Framework Recommends Federal Preemption of State AI Laws, No New Regulatory Body

By Hector Herrera | April 17, 2026

The Trump Administration released its National Policy Framework for Artificial Intelligence on March 20, recommending that Congress pass legislation to override state AI laws the administration deems an "undue burden" on development — and explicitly ruling out any new federal AI regulatory agency. The framework sets the White House's negotiating position as Congress drafts competing AI bills from both parties this spring.

This is not a neutral document. It is a policy roadmap that prioritizes AI development over AI regulation, and it puts the federal government directly in opposition to states like New York, California, and Colorado that have passed or are advancing their own AI oversight laws.

Context

The Biden Administration's 2023 Executive Order on AI established voluntary commitments, safety evaluations, and watermarking standards — a framework built around consensus and cooperation with the AI industry. The Trump Administration revoked that order in January 2025 and has been signaling since then that its approach would be more permissive.

The March 20 framework is the first comprehensive statement of what the administration actually wants Congress to do. It arrives as the Senate Commerce Committee and House Energy and Commerce Committee are both actively drafting AI legislation, and as states have passed more than 700 AI-related bills in 2025 alone — creating a patchwork that AI companies say makes national development strategies difficult.

What the Framework Says

On state preemption: The framework recommends Congress pass legislation that preempts state AI laws found to constitute undue burdens on interstate commerce or AI development. It does not specify a blanket preemption of all state AI law — but the framing targets laws with compliance costs, liability exposure, or operational restrictions on AI systems used across state lines. New York's RAISE Act, California's AI procurement rules, and similar statutes are in scope.

On regulatory architecture: The framework explicitly rejects creating a new federal AI regulatory body — no equivalent of the SEC for AI, no new agency modeled on the UK's AI Safety Institute. Instead, it recommends governing AI through existing agencies: the FTC for consumer protection, NIST for standards, sector-specific regulators (FDA, FAA, FERC) for high-risk applications in their domains.

On priorities: The framework's two named federal priorities are child safety and free speech. On child safety, it supports affirmative obligations on AI systems that interact with minors. On free speech, it recommends prohibiting government-funded AI tools from being used to restrict or score lawful speech — a provision aimed at content moderation systems and AI-assisted government communications.

On liability: The framework explicitly does not recommend federal liability rules for AI developers or deployers. This is a significant omission. Most state AI bills that have passed or advanced include some form of liability for algorithmic harm; the White House framework would preempt those rules without replacing them with federal alternatives.

What This Means

For AI companies: This framework is about as favorable a federal position as the industry could have expected. No new agency, no federal liability, and an active push to clear the state-level compliance maze. The main uncertainty for companies is whether Congress will actually pass preemption legislation — a harder political lift than the White House framework implies.

For states: New York's RAISE Act signed on March 27 was a direct response to — and direct challenge to — this framework. States that have invested in AI oversight legislation are not going to voluntarily surrender it. The legal and political fight over preemption will play out over years, not months.

For consumers: The framework offers no federal floor on AI safety, transparency, or liability. If preemption passes and state laws are struck down, the consumer protection baseline for AI systems reverts to existing FTC unfair trade practice authority — a standard not designed for algorithmic systems and unlikely to produce AI-specific enforcement at scale.

For the regulatory landscape: The "use existing agencies" approach has a credibility problem. The FDA has authority over AI in medical devices but not diagnostic AI tools. The FAA regulates AI in aircraft but not in air traffic management software sold to airports. The gaps in existing authority are well-documented, and routing AI governance through agencies with jurisdictional limits means those gaps become permanent unless Congress separately expands each agency's mandate.

The Legislative Landscape

Both parties are drafting AI bills. Republican versions are expected to align closely with this framework — preemption, no new agency, child safety focus. Democratic versions are expected to include liability rules, mandatory safety testing, and federal transparency requirements that would give the framework's child-safety goals regulatory teeth.

Whether any of these bills passes the Senate's 60-vote threshold is a separate question. The most likely path to federal AI legislation in 2026 is a narrow, bipartisan bill focused on child safety or national security uses — not a comprehensive framework bill.

What to Watch

Watch the Senate Commerce Committee markup. If the committee adopts preemption language in a bipartisan bill, states like New York face a real legal threat to their AI oversight regimes. If preemption stays in the partisan lane, the state-level patchwork continues and grows.

Source: White House National Policy Framework for Artificial Intelligence, March 20, 2026