White House Releases “Policy Framework for Artificial Intelligenc

This morning, the White House released a four-page “National Policy Framework for Artificial Intelligence,”[1] which builds on several months of the Trump Administration’s policy statements on the roles of state and federal governments in AI regulation.[2]

This comes only two days after Senator Marsha Blackburn (R-TN) released a sweeping 300-page discussion draft of the “TRUMP AMERICA AI Act,”[3] which seeks to codify President Trump’s Executive Orders on AI. Despite sharing priorities, the two documents disagree on copyright, developer liability, and Section 230.

The Policy Framework’s Seven Areas of Focus

The White House’s Policy Framework covers seven categories, but four carry the most weight for companies deploying or developing AI:

Federal preemption of state AI laws. States would be barred from regulating AI development, period. They could not impose burdens on AI use for activities that would be lawful without AI, and they could not hold developers liable for how third parties misuse their models. But the framework carves out broad exceptions: states keep authority over child safety, fraud, consumer protection, zoning, and their own government’s procurement of AI. Drawing the line between “AI development” (preempted) and “general consumer protection” (preserved) will be the central fight if this becomes law.

Copyright deferred to the courts. The administration says it believes training AI on copyrighted material is lawful, but rather than codify that position, it tells Congress to stay out and let judges resolve the fair use question. It also supports creating collective licensing frameworks so rights holders can negotiate with AI companies without triggering antitrust liability. This directly contradicts Blackburn’s bill, which would declare AI training on copyrighted works categorically outside fair use.

Child safety as the anchor. Age-assurance requirements, platform features to reduce exploitation and self-harm risks, and extension of existing child privacy protections to AI systems. This is the one area with real bipartisan support, and the framework deliberately preserves state authority here, including over AI-generated child sexual abuse material.

No new federal AI agency. Instead of a centralized regulator, the framework routes oversight through existing agencies with relevant expertise (SEC for financial AI, FDA for health, FTC for consumer issues). Congress is also told to create regulatory sandboxes, though the framework says nothing about which agency would run them or how they interact with existing rules.

The remaining three sections cover energy (data centers should pay for their own power, not ratepayers), free speech (a cause of action against government censorship via AI platforms), and workforce (study displacement, add AI to existing training programs).

Does this Solve the Open Questions of Federal Preemption?

In a word: “no,” and the diversity of state-level AI regulations under consideration and entering effect will only complicate the analysis.

Although the preemption concept sounds clean in this four-page summary, the boundary between preempted “AI development” regulation and preserved “general applicability” laws is undefined and untested.[4] Colorado’s AI Act, which takes effect June 30, imposes obligations on both developers and deployers of high-risk systems. Does it regulate “AI development” or enforce consumer protection? The framework doesn’t answer that question, and newer state AI regulations are also including private causes of action that may be a response to federal preemption tactics.[5] Congress has already rejected preemption twice: once in the One Big Beautiful Bill Act (stripped by a 99-1 Senate vote) and once in the FY26 defense authorization bill.

For now, this framework changes nothing about existing compliance obligations. But it’s a strong signal of the Trump Administration’s intent, especially with other federal agency deliverables related to state AI laws due this month. The FTC and Secretary of Commerce were instructed to provide assessments of “onerous” state AI regulations, which may serve as a roadmap for the Department of Justice’s AI Litigation Task Force. Whether this Policy Framework plays out in Congress or the courts, we’ll be tracking both paths.