On June 17, Axios reported that Representative Lori Trahan’s federal AI regulation plan was still taking fire from progressives. The 269-page discussion draft she released on June 4 with Republican Jay Obernolte — formally the Great American Artificial Intelligence Act of 2026 — is the most substantive bipartisan AI legislation this Congress has produced, and the last realistic window for federal AI rules before the August recess and the midterm elections.
The draft’s core is a bargain: the largest AI developers accept mandatory safety plans and periodic third-party audits, and in exchange states lose, for three years, the power to pass new laws governing how models are built. The backlash arrived within hours — and notably, from Trahan’s own party.
What Is Inside the 269 Pages
Under the draft, the most powerful AI developers must create and implement plans to address the potentially catastrophic risks of their advanced models — above all the prospect of new systems supercharging cybersecurity threats. Third-party auditors verify compliance on a semi-annual schedule, and developers must keep the government informed about frontier model development as it happens.
The bill also writes an executive-branch initiative into statute: a Center for AI Standards and Innovation (CAISI) inside the Commerce Department’s NIST, funded at $300 million to evaluate frontier models for the next three years. Trump’s executive order earlier that same week had tried to stand up the same office, but without congressional appropriations it was an empty shell. The draft is the funding vehicle. Four additional cross-party co-sponsors — Suhas Subramanyam, Scott Franklin, Scott Peters, and Erin Houchin — signed on at release.
The Preemption Design: Development Yes, Deployment No
The most contested clause preempts, for three years, state laws that specifically regulate AI model development — how systems are trained, built, and weighted — with the preemption sunsetting after three years. State laws governing the deployment or use of AI systems are explicitly excluded: civil rights enforcement, labor protections, copyright, child sexual abuse material prohibitions, and consumer privacy all remain with the states.
The sponsors describe that split as a deliberate compromise. But the FAQ from Trahan’s own office names the casualties, starting with California’s AB 2013, which requires model developers to publicly post training-data summaries. That is exactly the kind of development-side transparency law the draft would freeze.
“Turning the Floor Into a Ceiling”: The Backlash
TechTimes described near-universal rejection landing within hours of release — from labor unions, consumer advocates, and even a formal House Democratic commission. Politico reported that state lawmakers in Massachusetts and New York warned Trahan directly against preempting their ability to regulate AI developers.
Brad Carson, president of Americans for Responsible Innovation, framed the sharpest critique: in its current form the bill would turn “the current floor on state AI legislation” into a ceiling, and preempting state authority would be “a generational mistake.” For a discussion draft — a document that exists precisely to absorb criticism before formal introduction — the breadth of the opening rejection was remarkable: the coalition that normally backs AI safety legislation split over whether any federal floor is worth trading away state authority for three years. By June 17, per Axios, that progressive pressure had not let up.
What It Means for Developers and Product Teams
Three practical readings. Timing: the draft is open for stakeholder comment before formal introduction, with feedback directed to GAAIA@mail.house.gov; if nothing coheres before the August recess, the post-midterm Congress will be harder terrain, not easier. Structure: if passed as drafted, model development would face a single federal standard instead of fifty accumulating state regimes — which is precisely why the largest labs favor it. Status quo: until passage, nothing changes — state AI laws keep taking effect and multiplying, and compliance teams keep mapping obligations state by state. One more wrinkle is the calendar itself: a bill that misses the August recess does not simply carry over — a post-midterm Congress may restart from a different baseline, or not restart at all. The federal-versus-state standoff also extends the tightening policy tone we flagged in our opening outlook for 2026.
Sources
- House unveils AI draft that would preempt state laws — Politico (via Yahoo News)
- New Bipartisan Legislation Takes a Big Step Forward in Restricting State Regulation of AI — Gizmodo
- Federal AI Regulation Bill Freezes State Consumer Protections for Three Years, Sparks Revolt — TechTimes
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.