Three provisions. Three distinct compliance audiences. One discussion draft that’s reshaping how corporate legal teams should be thinking about US AI governance in 2026.
The Great American AI Act of 2026, described as bipartisan, though sponsor lists haven’t been confirmed, is circulating in Congress as a discussion draft. It doesn’t have a bill number yet. It hasn’t been introduced as formal legislation. But its provisions are specific enough that compliance teams can’t afford to treat it as noise, according to reporting on the draft’s circulating text.
The first provision would allocate $100 million per year for a federal AI standards body. That figure, corroborated by coverage from early June, would create the institutional infrastructure that the voluntary EO framework doesn’t establish. Right now, there’s no single federal body with authority to set binding AI standards. The GAAIA would change that, if enacted.
Who This Affects
The second provision would freeze state AI development rules for approximately three years. That’s the one state AGs and multi-state operators are watching most closely. Colorado, Connecticut, Illinois, and New York all have active or pending AI compliance frameworks. A federal preemption clause wouldn’t void those laws retroactively, but it would prevent new state rules from taking effect during the freeze window, creating a period of relative federal primacy. The exact duration is corroborated by prior reporting but hasn’t been verified against the draft’s legislative text, treat the three-year figure as reported, not confirmed.
The third provision would create new employer obligations under the WARN Act for AI-driven layoffs. Specific headcount thresholds and notice timelines aren’t confirmed in available material, so don’t build policy around exact figures yet. What’s confirmed: a WARN Act provision exists in the draft, and it would create obligations for employers who use AI systems in workforce reduction decisions. HR teams and employment counsel should be tracking this alongside the state WARN Act amendments already moving in several states.
The context matters here
The Trump administration signed a voluntary access Executive Order on approximately June 2, establishing federal access to frontier AI models without imposing binding requirements. NSPM-11, signed June 5, addressed military AI adoption. The GAAIA isn’t a direct legislative response to those actions in any confirmed causal sense, it was circulating in the same window, not triggered by them. Framing it as a “response” to the EO is editorial shorthand, not a verified fact.
Verification
Partial Congressional reporting via press coverage, no primary source text verified All GAAIA provisions ($100M/yr, 3-year freeze, WARN Act trigger) are corroborated by prior reporting but unverified against draft legislative text. Treat as reported, not confirmed.The catch is that discussion drafts rarely survive markup intact. The $100M figure, the three-year preemption window, and the WARN Act thresholds are all negotiating positions, not settled law. The real question is which provisions have coalition support broad enough to survive the process, and right now, the preemption clause is the one most likely to draw the hardest opposition from state-level interests.
Don’t expect a quick path to enactment. But the combination of a voluntary executive framework and a binding legislative draft in the same two-week window tells compliance teams something specific: the range of possible US AI governance outcomes has narrowed. Build your monitoring program around the provisions most likely to survive, not the draft as written.