techpolicy.press web signal

New Bipartisan AI Draft Targets Frontier Labs, Freezes State Rules

regulation ai-regulation federal-legislation

TL;DR

  • Representatives Obernolte and Trahan's 269-page discussion draft would impose binding compliance requirements on AI companies with over $500 million in annual revenue.
  • The bill would preempt state AI development laws for three years, through December 2029, while preserving state authority over AI deployment and use.
  • The House AI Commission co-chairs have already said the draft cannot serve as a basis for productive dialogue, signaling a difficult path forward.

A 269-page bipartisan discussion draft from Representatives Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.), released on June 4, would be the first attempt to put binding federal rules on the companies building America's most powerful AI systems. Tech Policy Press has published a detailed analysis of what the bill, called the Great American Artificial Intelligence Act of 2026, actually proposes.

The bill targets "large frontier developers," meaning companies with more than $500 million in annual revenue that have trained a frontier model, capturing OpenAI, Anthropic, Google, Meta, and xAI. Those companies would face mandatory transparency requirements, independent auditing, and semi-annual compliance verification by an Independent Verification Organization licensed through NIST's Center for AI Standards and Innovation. The bill also proposes boosting CAISI's annual budget from $15 million to $100 million, creates whistleblower protections including reinstatement, double back pay, and attorney's fees, and updates WARN Act requirements to cover mass layoffs caused by AI. Civil penalties for non-compliance could reach $1 million per violation per day.

The most politically charged element is the preemption clause. The bill would freeze state laws regulating AI model development for three years, through December 2029. Rep. Trahan has characterized the scope as narrow, saying the preemption covers "frontier model development, not products, not deployment, not use," with states retaining authority over how AI is actually used within their borders. The Business Software Alliance and the Information Technology Industry Council have expressed support; Public Citizen, Public Knowledge, and the AFL-CIO have pushed back on the preemption provisions.

The honest caveat is that this faces steep political obstacles. The House AI Commission's co-chairs have already said the draft "cannot serve as the basis for productive dialogue," and it remains a discussion draft, not a formal bill with a hearing scheduled. What the draft also leaves open is how enforcement would function during the period before the Independent Verification Organization system is actually licensed and running, and what happens to state laws already on the books during any transition.

For practitioners, the practical value of a federal framework is regulatory clarity: one standard replacing a patchwork of state rules that are multiplying and pointing in different directions. Whether the version that might eventually pass retains enough compliance teeth to make that tradeoff worthwhile is the question to track.

Shared on Bluesky by 2 AI experts