Federal agencies sued over secret AI safety framework

Four U.S. federal agencies are facing a lawsuit from the nonprofit Protect Democracy regarding a secretive framework used to test frontier AI models. The organization alleges that the Trump administration has failed to provide transparency regarding how these models are evaluated, who participates in the process, and what legal authority governs these reviews. Protect Democracy argues that this lack of oversight could allow the government to manipulate AI access and influence which models reach the public. The lawsuit targets the Office of the National Cyber Director, the Office of Science and Technology Policy, and the Departments of Treasury and Commerce. While the White House maintains that certain details must remain classified for national security reasons, the plaintiffs are seeking a court order to disclose all non-classified records by September 30 to ensure public accountability and congressional oversight amidst growing concerns over potential corruption and political influence in AI regulation.