What is the White House AI legislative framework?

The White House AI legislative framework is the longer-form, draft-statute-style document that accompanies the March 20, 2026 framework release. Where the framework release is the political statement of priorities, the legislative framework is the text that congressional staff can read to draft actual bills. It is published as a PDF on the White House site and was issued alongside the release as the technical companion.

What it actually contains:

The document restates the six priorities from the framework release but goes deeper on each one, providing statutory language proposals and implementation details. The expanded sections include:

• Child safety provisions. Specific recommendations on age-assurance mechanisms, parental-consent flows, and the kinds of AI-generated content that should require explicit labeling. The framework recommends a federal floor that would preempt weaker state child-safety laws while allowing states to impose stricter ones in some areas.
• Intellectual property and training data. The framework proposes disclosure rules for AI training datasets, including what categories of content are included and rights-holder notification. This is the most contested section — AI labs have argued broad training data access is essential; rights-holders (writers, artists, journalists) have argued for compensation and consent.
• Workforce and education. Specific funding recommendations for retraining programs, apprenticeship pathways, and K-12 AI literacy. The framework frames these as complements to the innovation priorities — the assumption being that AI displacement will require substantial workforce transition support.
• Federal preemption. Detailed language about which state AI laws would be preempted and which would survive. The preemption section is the most legally contested part of the document — the Congressional Research Service has published analysis noting that federal preemption of state AI law is more legally complex than preemption in earlier technology areas, because AI touches so many existing regulatory domains (privacy, civil rights, employment, consumer protection).

What the framework does not contain:

• Specific penalties. The framework describes what Congress should regulate but does not specify fines or enforcement mechanisms — those are left to congressional drafting.
• Preemption of all state AI law. The framework preempts specific categories of state law (those that “impose undue burdens” on AI development or “require alterations to the truthful outputs of AI models”), not state AI law generally. State laws on AI-driven employment discrimination, AI in criminal justice, and AI in healthcare are typically preserved.
• Implementation timelines. The framework is a recommendation; the actual implementation depends on congressional action and agency rulemaking, neither of which has hard deadlines in the document.

What this means in practice:

• For Congress. The document is a starting point for AI bills that have been in committee throughout 2025 and 2026. Several of its provisions — particularly the IP and preemption sections — are likely to be the focus of the next round of congressional AI hearings.
• For companies. The framework signals what compliance will look like if Congress follows the recommendations. A federal floor simplifies multi-state compliance; a federal preemption of state disclosure rules simplifies product design but limits flexibility for state-specific markets.
• For states. California, Colorado, Illinois, and several other states have signaled they intend to defend their authority to set AI rules regardless of federal preemption. Expect litigation over the scope of federal preemption if Congress enacts the framework’s preemption provisions.

The what-is-the-white-house-ai-framework answer covers the policy priorities at a higher level. This answer is the technical companion, covering what the legislative recommendations actually say and how they would interact with existing state law.