The Great American Artificial Intelligence Act of 2026 is still a discussion draft, and by Covington’s read there’ll be no floor action on it as drafted in the rest of the 119th Congress. Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) released the nearly 270-page text on June 4. Two months later, the bill sits in a bipartisan crossfire that isn’t resolving.

GAAIA would preempt state laws “specifically regulating the development” of AI models for three years, codify the Center for AI Standards and Innovation at NIST, and add transparency, audit, whistleblower, and cyber provisions. Trahan has said she signed on after Anthropic’s Mythos model demonstrated unprecedented cyber capabilities, which locates the bill’s frontier-model logic in a specific incident rather than an abstract fear.

The politics are the story. House Democrats view the preemption as too broad; House Republicans and the White House want it broader still. From the left, ARI president Brad Carson said the draft “takes the current floor on state AI legislation and turns it into a federal ceiling,” and Rep. Ted Lieu (D-CA) argued it fails civil rights, labor, and watchdog groups. Mintz notes most observers now expect the stalemate to continue.

The states aren’t waiting. On July 1, Colorado Gov. Jared Polis signed the Chatbot Safety Act (H.B. 26-1263), effective January 1, 2027. Illinois S.B. 315 adds frontier-model rules. And because GAAIA’s preemption targets development, deployment-stage state regimes would survive it anyway.

That’s the environment vendors serving small and mid-size businesses, Glean, Dust, and LemonLime among them, are actually building compliance programs against. Washington is drafting a ceiling nobody will vote on. The floor is being poured in Denver and Springfield.

Sources

Sources