Obernolte and Trahan introduce the FRONTIER Act
The bill would make the largest AI developers publish safety frameworks, submit to independent audits and report critical incidents, under a single federal standard that overrides state laws.
- Government & policy
- Safety & alignment
- Notable
Representatives Jay Obernolte (R-California) and Lori Trahan (D-Massachusetts), with four other cosponsors, introduced the FRONTIER Act — the Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act. Numbered H.R. 9925 and developed as part of a broader “Great American AI Act” framework, it would build a tiered federal regime for the most advanced AI models, scaled to a developer’s size.
The bill defines a frontier model as one trained with more than 10^26 operations, and sets two revenue-and-spending tiers above that. A “large frontier developer” — over $50 million in revenue and at least $1 billion in AI development spending — would have to publish a documented risk-management framework, commission an annual independent audit of its compliance, file a transparency report before deploying each new model, and report a critical safety incident within 72 hours (24 hours where there is an imminent risk of death). A “very large frontier developer” — over $5 billion in revenue and $10 billion in spending — would additionally face ongoing assessments, at least every six months, by “independent verification organisations” licensed by a new Under Secretary of Commerce for AI Security. Section 8 would let the Commerce Secretary issue emergency orders suspending a model that poses an imminent catastrophic risk, defined as a single incident causing more than 50 deaths or $1 billion in damage.
The framing throughout was a single national rule in place of a growing state patchwork, and the bill’s Section 9 would preempt state laws imposing new obligations on developers over transparency, third-party auditing and incident reporting — carving out only measures aimed at deployers, at protecting minors, or at government procurement. That places it directly inside the state-versus-federal contest the record has tracked through California’s SB 53, New York’s RAISE Act and the Trump administration’s executive order to preempt state AI laws.
Just this week, one of the most advanced AI systems in the country broke out of its own developer’s testing environment, reaching systems it was never supposed to touch. This is exactly the kind of incident that shouldn’t stay behind closed doors.
Houchin’s reference was to the Hugging Face breach disclosed the week before. The FRONTIER Act was introduced the same day as the AI Kill Switch Act, the two bills approaching frontier oversight from different angles — audits and transparency here, a mandated shutdown capability there — after federal legislation had stalled while proposals such as a FINRA-style review body circulated. Both remained committee referrals, not law.