Federal pre-emption: because nothing says ‘democracy’ like telling fifty states to sit down, be quiet, and let the national competitiveness narrative do its thing.
Washington wants a single national AI rulebook. The states want to keep writing their own. And Big Tech would like everyone to stop touching the controls while the machine is accelerating.
What happened
On March 20, the White House released a “National AI Legislative Framework” and accompanying legislative recommendations arguing that the framework must be applied uniformly across the U.S.—explicitly warning that “a patchwork of conflicting state laws would undermine American innovation.” (White House release)
The PDF itself goes further, recommending Congress “preempt state AI laws that impose undue burdens” to avoid “fifty discordant” regimes, and arguing states “should not be permitted to regulate AI development” because it is “inherently interstate” with “foreign policy and national security implications.” (National Policy Framework for Artificial Intelligence — Legislative Recommendations, March 2026, p.4)
Meanwhile, reporting from NPR/WBUR captures the other half of the tension: Congress isn’t moving fast, so state lawmakers are stepping in—creating exactly the patchwork the federal framework says it wants to prevent. (WBUR / Here & Now (NPR))
The non-obvious angle: pre-emption is an industrial policy move
This isn’t just a legal turf war. It’s market-structure engineering.
If you’re building and deploying AI at hyperscaler scale, the scariest scenario is not “regulation exists.” It’s “regulation exists in fifty incompatible versions,” each with its own disclosure rules, liability triggers, and compliance mechanics. That’s not a safety policy. That’s a tax on speed—and speed is the only thing the AI industry is actually optimized for.
A Stakes Map: who wins, who loses
Winners
Big deployers (and their lawyers) — A single federal standard reduces compliance fragmentation. “Uniformity” is a softer word for “we can ship.”
Federal executive agencies — Pre-emption shifts the center of gravity toward federal policy, federal enforcement priorities, and federal narratives about “dominance.”
Whoever can define ‘undue burdens’ — The PDF’s language is broad by design. The fight will be over what counts as “undue,” and that fight is where lobbying becomes law.
Losers
States trying to move first — Many state AI bills are attempts to fill a vacuum. Pre-emption turns that energy into a compliance dead-end.
Consumers who liked the idea of local experimentation — State policy can be messy, but it’s also how the U.S. prototypes governance. A single federal rulebook tends to reflect whichever interests got to the drafting table first.
The carve-outs that matter (and the loopholes they create)
The PDF explicitly says federal pre-emption should not erase all state authority. It preserves traditional “police powers” (fraud/consumer protection/child protection), state zoning authority for infrastructure placement, and “requirements governing a state’s own use of AI… through procurement.” (March 2026 recommendations, p.4)
Translation: the states may not be able to regulate AI development broadly, but they can still fight via: (1) enforcement of general laws; (2) infrastructure siting battles; and (3) procurement rules for their own agencies. Expect the real action to migrate into those channels, because politics abhors a vacuum almost as much as a lobbyist does.
The Singularity Soup Take
“Uniformity” is being sold as public clarity. It’s also a competitive weapon. When Washington says ‘no patchwork,’ what Big Tech hears is ‘we get one battlefield instead of fifty.’ The weird part is pretending this is about soothing worried parents, rather than clearing the runway for the biggest systems to scale without tripping over state-level guardrails.
What to Watch
1) The definition of “undue burdens” (it will decide what survives). 2) Which state laws get targeted first (child-safety carve-outs vs broader model governance). 3) Whether infrastructure siting and grid cost allocation become the de-facto “AI regulation” battleground as states use zoning and permits as leverage.
Sources
The White House — “President Donald J. Trump Unveils National AI Legislative Framework”
The White House (PDF) — “National Policy Framework for Artificial Intelligence — Legislative Recommendations (March 2026)” (see p.4 for pre-emption language and carve-outs)
WBUR / Here & Now (NPR) — “Congress isn't regulating AI, so states are stepping up”