US Lawmakers Introduce FRONTIER Act to Require Independent Audits of Frontier AI Models

A Bipartisan Bill in Congress

On July 23, 2026, US Representatives Jay Obernolte, a Republican from California, and Lori Trahan, a Democrat from Massachusetts, introduced the FRONTIER Act in the House of Representatives, short for the Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act. According to Obernolte's official press release, the bill carries bipartisan cosponsorship from Representatives Scott Franklin of Florida, Scott Peters of California, Erin Houchin of Indiana and Suhas Subramanyan of Virginia.

What the Bill Would Require

The legislation sets up a national, risk based regulatory framework aimed specifically at the most advanced AI systems, known as frontier models. Under the text, large AI developers would have to publish transparency reports when deploying frontier models, maintain formal risk management frameworks, report critical safety incidents to federal regulators and undergo audits carried out by independent third party verification organizations, rather than simply self certifying compliance.

One Piece of a Larger Bill

The FRONTIER Act grows out of a much broader proposal. Obernolte and Trahan had released a bipartisan discussion draft called the Great American AI Act on June 4, 2026, to gather feedback from experts, industry and the public ahead of formal introduction. As Tech Policy Press has reported, that wider draft remains stuck on a preemption fight over whether a federal AI law should override individual states' own AI statutes.

The Preemption Standoff

Per Tech Policy Press's analysis, House Democrats oppose broad preemption of frontier model rules, worried it would gut existing state protections, while House Republicans counter that the proposed preemption does not go far enough to cover the rest of the AI ecosystem. By carving the third party verification and incident reporting piece into a standalone bill, Obernolte and Trahan are trying to move part of the framework forward while that harder argument continues separately.

Industry Reaction and Cybersecurity Framing

A release distributed via PRNewswire shows Fathom applauding the introduction of the FRONTIER Act as a meaningful step toward regulatory clarity for the sector. Cybersecurity Dive's coverage notes that the bill's combination of independent audits and mandatory incident reporting connects to cybersecurity oversight structures already used by the federal government, and raises questions about how open weight, or open source, models would be treated under the regime.

Why This Matters for Brazilian Agencies and SMBs

The FRONTIER Act's obligations fall on large US developers of frontier models, such as OpenAI, Anthropic and Google, not on companies abroad that use those tools. Still, a mandatory incident reporting and independent audit regime tends to put more public information into circulation about real safety failures and model limitations, information Brazilian agencies and SMBs can factor into decisions about which foundation model to build on. The bill's structure also previews the kind of transparency requirements, model cards, audits, incident disclosure, that Brazil's own PL 2338/2023 AI framework, still moving through the Câmara dos Deputados, may converge toward. It is worth watching the FRONTIER Act as an early indicator of future Brazilian compliance obligations.