Here’s the intro from this Cooley Alert penned by Sean Quinn, Michael Egan, Janet Kim, Bobby Ghajar, Jude Lauter, Tristan Lockwood, Adam Silow and Rebecca Kahn:
“On March 20, 2026, the White House released its National Policy Framework for Artificial Intelligence (“the Framework”) outlining the administration’s recommended federal approach to AI regulation. The Framework is the most concrete statement yet of where the administration wants Congress to take federal AI policy. If Congress adopts this approach, it would reshape the US AI regulatory landscape, significantly affecting how companies navigate an already complex web of state, federal and global obligations.
The Framework is a follow-through from the December 11, 2025 executive order (EO) “Ensuring a National Policy Framework for Artificial Intelligence,” which we discussed in this December 12 alert. This new set of recommendations includes many elements that have previously been advocated for by the Trump administration, including preemption of some state AI laws, which was first included in early versions of the “One Big Beautiful Bill” legislation last year.
The Framework encourages Congress to pass laws protecting children and their data in the AI context, but leaves states with the ability to enforce their generally applicable child protection laws. Importantly, the Framework specifically states that Congress should not preempt consumer protection laws that may apply to AI, which is one of the primary bases on which states are regulating the consumer-facing AI industry.
In other areas, the Framework generally supports the AI industry, promotes AI adoption through measures like regulatory sandboxes, and discourages new regulatory regimes or agencies specific to AI. The Framework also states that the Trump administration believes training on copyrighted material does not violate copyright laws but recommends leaving the courts to resolve the issue in specific cases and contexts. It generally encourages Congress to enable a nonmandatory licensing framework for training data.”
Authored by
Broc Romanek