← Back to feed

Six lawmakers from both parties released a 269-page discussion draft called the Great American Artificial Intelligence Act on June 4. It is the most structurally complete attempt at federal AI legislation since the current administration took office, and it landed with immediate controversy.

What the Bill Does

The draft sets a revenue threshold: companies earning more than $500 million annually qualify as “large frontier developers”, and it imposes four obligations on them:

1. Publish a frontier AI framework documenting catastrophic risk assessment, model weight security, incident handling, and release timelines
2. Report critical safety incidents within 15 days; imminent risks within 24 hours
3. Retain an Independent Verification Organization licensed through CAISI for semi-annual compliance audits
4. Submit audit results to CAISI

The draft formally establishes CAISI in statute (the AI Standards and Innovation center that originated as the Biden-era AI Safety Institute and was rebranded by Commerce in 2025) and funds it at $300 million over FY2027-2029. CAISI’s authorities sunset after three years.

The State Preemption Clause

The most contested provision: the bill preempts state laws “specifically regulating the development” of AI models for a three-year window. Deployment rules remain with states.

The draft document explicitly names California AB 2013 (training data summaries) and portions of SB 942 (content watermarking) as preempted. Frontier safety laws in California, New York, and Illinois would be “federalized” under the bill’s framing.

Reps. Lori Trahan (D-MA) and Jay Obernolte (R-CA) co-introduced alongside four others. Trahan framed it as protection-first: “The threats AI poses to our national security, our safety, and our workforce are here and growing.” Houchin (R-IN) framed it as competitiveness: “America should lead the world in artificial intelligence, not regulate ourselves into falling behind China through a patchwork of fifty different state laws.”

What “Catastrophic Risk” Means in the Bill

The draft defines catastrophic risk as “a foreseeable and material risk” of more than 50 deaths or more than $1 billion in property damage. Developers must identify internal thresholds they use to determine when a model crosses that line and disclose those thresholds publicly.

That definition is narrower than California’s existing frontier model safety law and narrower than what AI safety advocates had pushed for. It will not capture harms that are diffuse, economic, or below the 50-person threshold.

Immediate Pushback

The bill drew organized criticism within hours. Americans for Responsible Innovation president Brad Carson: “This bill takes the current floor on state AI legislation and turns it into a federal ceiling, preventing state lawmakers from addressing emerging AI harms in an era of fast-moving technology. Tying their hands would be a generational mistake.”

Democrats objected to the preemption language even while supporting the bill’s intent. Republicans flagged that regulation at all could stifle innovation. The bill is explicitly framed as a discussion draft to solicit public feedback before formal introduction.

The Numbers

Provision	Detail
Revenue threshold	$500M annual
CAISI budget	$300M over FY2027-2029
State preemption window	3 years (development only)
Catastrophic risk floor	50+ deaths or $1B+ damage
Incident reporting	15 days (discovered), 24 hours (imminent)
Audit cadence	Semi-annual
CAISI sunset	3 years post-enactment

What Happens Next

This is a discussion draft, not an introduced bill. Formal introduction comes after the public comment period closes. Even then, the preemption clause has enough opposition from both flanks, safety advocates on the left and anti-regulation voices on the right, to require significant revision before floor consideration is realistic.

The practical effect of releasing the draft: every major AI developer in the country now knows exactly what a federal compliance regime would look like, and legal teams are mapping exposure against the $500M threshold.