Four US Agencies Sued Over Secret Frontier AI Testing Framework

Four US federal agencies are being sued over a secret framework the Trump administration uses to test frontier AI models before release.

Protect Democracy claims that “almost no details” have been disclosed about how the framework works, who participates, how companies are selected, or what legal authority supports the reviews. The nonprofit is seeking disclosure of the framework, participation terms, participant identities, access rules for frontier models, and related contracts.

The group also wants an injunction preventing agencies from withholding records that are not classified. The agencies named are the Office of the National Cyber Director, the Office of Science and Technology Policy, and the Departments of the Treasury and Commerce.

Protect Democracy argues that the government could influence which AI models are released and who can access them. It says the information is needed for congressional oversight and public accountability, and is seeking a court order requiring disclosure of the non-classified framework and related details by September 30.

The White House says some parts of the evaluation legitimately need to remain secret and that national security tests could contain classified information. A White House spokesperson said that unclassified information would not necessarily be broadcast publicly.

The lawsuit follows tensions between Anthropic, the maker of Claude, and the US Government after revisions to the availability of Mythos 5 and an attempt by President Trump to blacklist the company. US Representative Greg Casar (D-Tx.) has also criticized the administration’s AI safety record.