What happened
On July 23, 2026, Reps. Lori Trahan (D-MA) and Jay Obernolte (R-CA) introduced the FRONTIER Act, a revised version of last month's discussion draft, establishing a federal framework for governing advanced/frontier AI models. The bill requires large AI developers to publish safety-capability reports, adopt risk-mitigation frameworks, report critical safety incidents, and undergo independent third-party audits, with a federal private-assurance market for oversight. It gives the government authority to halt deployment of models posing imminent catastrophic risk, but leaves the initial release decision with developers. The bill preempts state laws specifically on frontier AI catastrophic-risk transparency, third-party auditing, and incident reporting (narrower and without the 3-year sunset clock of the prior draft), while carving out state authority over deployment and child-safety regulation.
Why it matters
This is the most detailed bipartisan House proposal yet for a comprehensive federal AI governance framework, and its preemption provisions would immediately override certain state AI catastrophic-risk laws (e.g., aspects of Illinois' and pending Massachusetts' AI audit statutes) upon passage, per critics including Americans for Responsible Innovation. It has no whistleblower protection and leaves risk/release decisions largely with AI labs.
Action needed
Frontier AI developers and state regulators should monitor for committee action; entities relying on state catastrophic-risk transparency/audit laws should assess preemption exposure if the bill advances.