Congress has a bill. It’s 269 pages long.
Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released the discussion draft of
the Great American Artificial Intelligence Act of 2026 on June 4. The draft proposes a three-year moratorium on state and local laws that regulate the
*development* of AI models, not a blanket federal takeover of the entire regulatory space. States would keep their authority over how AI systems are *used* within their borders. The
distinction matters enormously for what happens next.
That development-vs.-use line is the structural core of the proposal. Under the draft, a state
law requiring annual algorithmic audits during model training would be frozen. A state law
prohibiting discriminatory AI-driven hiring decisions by an employer would not. The bill’s
drafters are threading a specific needle: they want a single federal framework governing how AI
gets built, while leaving consumer and employment protections at the state level intact.
GAAIA: Initial Stakeholder Positions
The draft also includes a mandatory federal testing and evaluation requirement for frontier labs. The country’s most powerful AI developers would be required to submit their models for federal
vetting, a significant shift from the voluntary 30-day pre-release window in the June 2
Executive Order. Voluntary federal access and mandatory federal vetting are not the same thing,
and compliance teams at frontier labs now have two overlapping federal frameworks to track
simultaneously.
Consumer advocates aren’t waiting to respond. Public Citizen
argues the bill would strip states of authority to address algorithmic bias, fraud,
deepfakes, and employment discrimination without guaranteeing equivalent federal protections. That’s their framing as an advocacy organization, not a neutral legal assessment, but the
underlying concern is structurally legitimate. A three-year moratorium creates a protection gap
if federal rulemaking doesn’t move at the same pace as the state laws it displaces.
The legislative path is uncertain. This is a discussion draft, not introduced legislation, it
hasn’t been formally filed, referred to committee, or scheduled for markup. The Lexington
Institute called it “arguably the most important and consequential action Congress
has taken on AI, ever.” That’s one think tank’s view. Whether it survives a divided
Congress is a different question entirely.
Compliance Deadline
What to watch: The discussion draft stage is where bills get reshaped or killed. The
critical pressure points are state attorneys general who have already enacted AI laws, civil
society organizations building opposition coalitions, and frontier labs evaluating whether the
mandatory vetting requirement is a net positive or negative for their competitive position. Watch for formal introduction and committee referral, that’s the first real signal of whether
the Obernolte-Trahan framework has institutional support beyond its two sponsors.
TJS synthesis: The catch is this: companies operating under Illinois SB 315, Connecticut PA
26-15, or Colorado SB 26-189 can’t pause their compliance programs on the strength of a
discussion draft. Those state laws are live. The moratorium doesn’t exist yet and may never. Compliance teams should treat this bill as a signal worth monitoring, not a reason to slow down
work already underway. The real question is whether the mandatory frontier lab vetting provision
will draw more opposition from industry than the preemption provision draws from states. If it
does, the bill’s coalition fractures before it ever reaches a floor vote.