You live in a country where four states — Colorado, California, Utah, and Texas — have passed their own AI laws. All 50 states have introduced AI legislation over the past 18 months. Your state legislators looked at the AI boom and said, "Someone should probably write some rules." Reasonable.
But the White House disagrees.
On March 20, 2026, the Trump administration released its National Policy Framework for Artificial Intelligence — the first comprehensive federal AI legislative blueprint in U.S. history. Not a set of suggestions. A 30-page instruction manual telling Congress exactly how to regulate AI. The core message: do less, not more. The framework stems from Trump's December 2025 executive order and lays out seven "policy pillars." Let's read the fine print.
Pillar 1: Federal preemption — meaning federal law overrides state law. Congress should block states from regulating AI development, imposing liability on developers for third-party misuse, or "unduly burdening" lawful AI use. Those four state AI laws? The administration wants them scrapped. Every proposed state regulation? Dead on arrival.
Pillar 2: Copyright blessing. The framework declares that "training of AI models on copyrighted material does not violate copyright laws." Training, for the uninitiated, means feeding massive amounts of text, images, and code into an AI system so it learns patterns — like a student reading every book in a library, except the student never paid for any of them. The White House says courts should decide fair use, then advises Congress not to legislate. Translation: don't touch it while AI companies train on whatever they want.
Pillar 3: Voluntary licensing. A consolation prize for content creators. Congress should let rights holders form collective licensing groups to negotiate compensation from AI developers without antitrust liability — antitrust being the set of laws that normally prevent competitors from teaming up. "You can negotiate for money, but we've already declared you have no legal leverage." Elegant.
Pillar 4: No new regulator. No new federal AI agency. Instead, existing agencies handle their own turf — the FDA for health AI, the SEC for financial AI, and so on. The philosophy: don't build new infrastructure, trust what already exists. Whether those agencies have the expertise or budget for AI oversight is a question the framework politely ignores.
Pillar 5: Child safety. Age verification for AI platforms used by minors, parental controls, privacy protections. The one pillar designed to get bipartisan applause and newspaper headlines. Real provisions, conveniently positioned as the framework's public face.
Pillar 6: Infrastructure. Streamlined permitting so data centers — the warehouse-sized buildings full of servers that AI runs on — can generate their own power on-site. AI companies have been complaining about electricity constraints for two years. The administration listened.
Pillar 7: Free speech. A clause preventing the government from pressuring AI providers into suppressing "lawful expression," with proposed legal avenues for individuals to challenge federal censorship.
So what did the White House actually build here? A deregulation package wrapped in a child-safety bow.
The state preemption fight is the real battleground. Fifty Republican lawmakers sent a letter supporting preemption. Democrats on key committees are already pushing back, calling it "an effort to prevent the passage of measures holding the tech industry accountable." California — which has functioned as the country's de facto AI regulatory laboratory — would lose its ability to set standards. With the GOP's razor-thin House majority, turning this framework into legislation before the November midterms will be a knife fight.
The copyright position is equally explosive. The New York Times, Getty Images, and dozens of content creators have active lawsuits against AI companies over training data. The White House just publicly sided with the defendants — while claiming it's leaving the question to courts. That's a judge saying "I have no opinion" while wearing a team jersey.
The cost is clear. States lose their regulatory power. Content creators lose legal leverage. AI companies gain a federal shield. The child safety provisions are genuine, and they'll generate the bipartisan photo ops. Everything else reads like a wish list drafted in a Silicon Valley boardroom.
What does this mean for you? If you're a developer, designer, or anyone building with AI tools — the regulatory uncertainty isn't going away, but it's shifting from 50 state-level fights to one federal cage match. Your tools won't change tomorrow. The rules around them might change permanently.
Ten days later, the framework hasn't become law. It probably won't before midterms — the House math doesn't work. But it doesn't need to. This document sets the terms of debate for the next decade of American AI regulation. Those terms: build fast, regulate later, and if a state tries to slow you down, the feds will handle it. The EU streamlined their AI Act. The U.S. just tried to preempt its own states. Same planet, different regulatory universes.
