The Trump White House released a national AI legislative framework three weeks ago. Most of the coverage focused on the headline: the administration wants federal AI dominance. But Lawfare just published the most detailed legal analysis of what the framework actually says, and it is more aggressive than the headlines suggested.

The framework targets state AI laws in three specific ways. First: states cannot regulate AI development. That means laws like Californias SB 53, which requires large AI companies to publish safety frameworks and report critical incidents, would be preempted. Second: states cannot hold AI developers liable for what third parties do with their models. That takes direct aim at Colorados AI Act, which creates a duty of care for developers whose systems make consequential decisions. Third: states cannot impose extra requirements on AI-assisted decisions that would be lawful if made by a human without AI.

That third point is the most radical. Colorados AI Act requires businesses to conduct annual impact assessments and implement risk management programs when using AI for hiring, lending, housing, and healthcare. The White Houses argument: if a human can make those same decisions without those extra requirements, why should AI face a higher bar? It sounds reasonable on the surface. But it ignores that AI operates at a scale and speed that human decision-making never could. One biased algorithm can deny ten thousand loan applications before a human reviewer finishes their morning coffee.

Here is the part that makes this genuinely dangerous for state regulators: the framework has bipartisan support. House Speaker Mike Johnson and Senator Ted Cruz are backing it. But so is Senator Maria Cantwell, the ranking Democrat on the commerce committee, who said the framework identifies key areas to address. When both parties agree on deregulation, it usually passes.

Lawfare notes the framework does carve out some areas for states. They can still enforce laws of general applicability, including consumer protection, fraud prevention, and child safety. States can also maintain control over AI use in government procurement and critical infrastructure. And they can regulate AI in ways that align with federal law.

But the carve-outs are narrow and the preemptions are broad. White House adviser Michael Kratsios went further in a recent interview, suggesting the administration would also preempt state laws banning particular verticals. He was specifically reacting to New Yorks SB 7263, which would create liability for chatbots that practice law or medicine without a license. Nevada and Illinois have already enacted similar laws banning AI chatbots from performing unlicensed therapy.

The timing matters. Congress has failed to pass a single meaningful AI law. States have filled the vacuum with over 100 bills across dozens of legislatures. Companies like xAI are already suing states (Colorado, this week) to block enforcement. The framework gives Congress a roadmap to wipe the state-level slate clean before the 2026 midterms.

The bet is clear: the White House believes federal preemption is the price of maintaining American AI dominance against China. The risk is equally clear: if Congress preempts state laws before passing strong federal alternatives, the country could end up with no meaningful AI regulation at all. Given Congress has been unable to agree on the definition of AI, let alone how to regulate it, that outcome is not unlikely. It is the most probable one.