OpenAI Backs Federal Bill Requiring Outside Audits of Frontier AI Models

OpenAI has told lawmakers it supports a provision in a bipartisan House bill that would require the largest artificial intelligence developers to submit their safety practices to outside verification, the company's first explicit endorsement of a binding federal audit requirement for its own models.

Chris Lehane, OpenAI's chief global affairs officer, told reporters this week that during a Capitol Hill meeting he specifically discussed how the company views the bill's independent-review mechanism and made clear OpenAI can support it. The provision appears in the FRONTIER Act, introduced by Representatives Jay Obernolte, a California Republican, and Lori Trahan, a Massachusetts Democrat, along with four other co-sponsors from both parties.

The bill would create the first federal framework specifically for frontier AI systems, the largest and most capable models. It sets tiered obligations by company size: public safety frameworks, incident reporting to the Commerce Department, and, for what the bill calls "very large frontier developers" — those with more than $5 billion in annual revenue and at least $10 billion in AI development spending over three years — a requirement to retain a licensed Independent Verification Organization. That outside body would assess whether a company's stated safety practices actually reduce catastrophic risk and report its findings to Commerce every six months, according to the bill's sponsors.

OpenAI's position matters because the company, valued in the hundreds of billions of dollars, would likely qualify as a "very large" developer under the bill's thresholds and would be among the first companies subject to outside inspection of its safety claims. The endorsement follows a September 9 blog post in which the company said it wanted mandatory, capability-based national rules rather than a state-by-state patchwork, and comes as OpenAI, Anthropic and Google have reportedly held informal talks for weeks about a shared approach to AI safety oversight, according to people familiar with the discussions.

The shift is notable given how recently AI developers resisted binding external oversight, arguing that mandatory audits could slow innovation or expose proprietary methods. Lehane's comments, reported by outlets covering the Capitol Hill meeting, suggest the calculus has changed as momentum builds in Congress for some federal standard, with industry increasingly preferring a single national rule to the patchwork of state AI laws that has begun to emerge.

The FRONTIER Act still faces a long path before any vote. It must clear committee, attract broader sponsorship, and survive competing proposals, including narrower transparency bills and a separate push by some lawmakers for outright moratoria on certain AI applications. Critics on the right have raised concerns about creating a new regulatory body with power over private companies, while some AI-safety advocates say the bill's audit requirements are still too weak, applying only to a handful of the largest developers and leaving enforcement details vague.

What happens next will depend largely on whether other major AI labs follow OpenAI in publicly backing the independent-verification provision, and on whether the bill's sponsors can attract Senate interest before the current Congress adjourns. For now, OpenAI's endorsement gives the FRONTIER Act's authors their most significant industry validation yet.