Federal AI Preemption Sounds Simple But the Exceptions Won't Be.
A new framework from the Trump administration calls on Congress to broadly preempt state AI laws. Colorado’s new AI law offers an early example of the boundary disputes that could follow.
On March 20, the Trump administration released a Framework for Artificial Intelligence calling on Congress to “prevent a fragmented patchwork” of state AI laws by creating a “minimally burdensome” national standard. That standard would broadly preempt state regulation of AI development, while still leaving several important areas to the states, including consumer protection, fraud, child safety, zoning, and state procurement.
At a high level, that sounds like simplification: one national rule instead of fifty state ones. And for many clients, a federal rule with some exceptions may sound far better than the patchwork the framework is trying to avoid.
But that does not eliminate the hard part. In our federal system, lawyers will still need to sort out where broad federal preemption ends and surviving state authority begins.
The Colorado Question
Colorado’s AI Act takes effect June 30. It imposes obligations on certain developers and deployers of high-risk AI systems, including impact assessments and risk-management requirements. If Congress later passes a federal law preempting state regulation of AI development, does Colorado’s law get wiped away? Or does some or all of it survive because it fits within an area the states still control, such as consumer protection?
For many companies, one federal rule may be far better than fifty different state regimes. But even then, lawyers will still need to sort out where federal preemption ends and surviving state authority begins.
What to Ask
If you’re advising clients on AI governance or compliance, three questions:
Are you underestimating how hard it will be to figure out where the lines are between broad federal preemption and surviving state authority?
Which current state-law obligations could survive because they are framed as consumer protection, fraud, employment, or another traditional area of state authority?
Has Congress actually shown that it will enact broad federal preemption? Congress has previously considered and rejected that approach.
The Point
The administration is presenting uniform nationwide regulation as an answer to the state-law patchwork. If Congress enacts the kind of statute the administration is calling for, a new federal law may simplify things overall. But as Colorado’s new AI law suggests, one of the next challenges will be figuring out which parts of state law are preempted and which survive.
Go Deeper
For a detailed analysis of the framework’s seven focus areas and the unresolved preemption questions, see Jason Epstein and Jeffrey Kelly, The White House Releases National AI Legislative Framework (Nelson Mullins, Mar. 23, 2026).
On Our Radar
AI litigation is consolidating around product liability. Plaintiffs are increasingly framing AI harms as design defects and failure-to-warn claims rather than novel AI theories — and the EU’s revised Product Liability Directive, which treats software as a “product,” is likely to show up in U.S. complaints as persuasive authority. (K&L Gates, NLR)
Colorado’s AI Act rewrite is moving. The state’s AI Policy Working Group unanimously approved a proposed framework replacing the original bias-audit mandate with transparency and disclosure requirements. The current AI Act still takes effect June 30 unless the legislature acts. (Ballard Spahr)
GSA proposed new AI contract terms for federal suppliers. A proposed GSAR clause would require AI safeguards in federal supply schedule contracts — the first AI-specific contract provision for the government’s largest purchasing vehicle. (JD Supra)
Connecticut’s AG: existing laws already cover AI. Attorney General Tong issued an advisory clarifying that the state’s civil rights, privacy, consumer protection, and antitrust statutes apply to AI use — no new legislation needed. Oregon and New Jersey issued similar guidance in late 2024 and early 2025. (Sheppard Mullin, NLR)
Trying to make sense of all this? I help law firms and companies navigate the legal side of AI implementation. Email me at adam@lawsnap.com or click here to schedule a short call.