Federal agencies sued amid allegations Trump admin could be using AI safety framework to hide AI manipulation and corruption

Four federal agencies are facing lawsuits over a secretive framework for testing frontier AI models before they are released, according to Protect Democracy. The nonprofit claims that an overwhelming amount of details surrounding the framework remains undisclosed. Allegedly, the government is withholding information about who is involved, how models are evaluated, and the legal framework supporting these assessments.

Protect Democracy is demanding a court order by September 30th to compel the government to release the non-classified details of the framework. The group also seeks an injunction to stop agencies from obstructing the disclosure of non-classified records related to the framework. This legal action comes amid a dispute between AI company Anthropic and the US government, following revisions to the availability of Mythos 5 and President Trump's attempts to blacklist the company.

While the government contends that certain evaluation aspects should remain confidential, Protect Democracy argues that the secrecy undermines transparency, congressional oversight, and public accountability.