On March 20, 2026, the White House released its long-awaited National AI Policy Framework — a four-page document that attempts to replace fifty competing state AI laws with a single federal standard. If it sticks, it will be the most consequential piece of AI governance since the EU AI Act. This article was written March 11, based on Congressional testimony and leaked drafts; the final release confirmed all key elements.

The 50-State Problem

Before this framework, US businesses faced a genuine compliance maze. California required AI transparency. New York mandated algorithmic audits in hiring. Illinois regulated biometric data. Texas was drafting its own framework. For any company operating nationally, that meant up to fifty separate compliance tracks with different definitions, timelines, and penalties.

The White House framework’s central mechanism is federal pre-emption: the new single standard overrides state-level AI laws that conflict with federal policy.

Pre-Emption: Which State AI Laws Get Overridden

Pre-emption doesn’t mean all state laws vanish. The framework overrides laws where it establishes federal standards: children’s privacy and CSAM protections, AI-generated fraud and scam rules, IP rights in AI-generated content, and national security AI restrictions. State laws addressing areas the federal framework leaves silent may survive legal challenge.

EU AI Act Divergence

While the US moves toward a single light-touch federal standard, the EU AI Act takes full effect for high-risk AI on August 2, 2026. The two regimes are structurally incompatible for multinationals.

For multinationals: you cannot build one compliance system that satisfies both. EU operations need full risk assessments, audit trails, and human oversight documentation. US operations are largely self-regulated under a single federal standard.

David Sacks and the Nvidia China Export Question

AI Czar David Sacks has publicly argued that advanced Nvidia chips should be permitted to ship to China. The reasoning: if the US withholds chips, China accelerates Huawei’s indigenous chip program. Allow exports, and Chinese AI development remains dependent on American hardware. Republican Speaker Mike Johnson endorsed the framework broadly.

The “TRUMP AMERICA AI Act,” introduced March 23, consolidates AI regulation with children’s online safety provisions — combining two Republican priorities into one legislative package.

What Your Legal/Compliance Team Needs to Know