Four federal agencies have been sued over the secrecy surrounding the Trump administration’s AI safety review process, with plaintiffs demanding full transparency by September 30. A nonpartisan nonprofit, Protect Democracy, alleges that the government has not released any details about the framework used to evaluate frontier AI models for safety prior to release. The lawsuit seeks an order requiring officials to disclose all information related to the framework, including its text, terms of participation, and criteria for access to AI models. The complaint also asks for an injunction to prevent the improper withholding of unclassified records. 'The executive branch is now choosing which companies can release their products and which customers get access to a technology that could shape the future of not only American industry and national security, but economies and governments abroad as well,' Protect Democracy said in a statement. Source: arstechnica

Protect Democracy argues that the framework could be 'corrupt,' with officials from the Office of the National Cyber Director, the Office of Science and Technology Policy, the Treasury Department, and the Commerce Department potentially favoring AI firms that align with Trump’s interests. The group also points to a recent ruling that found it illegal for Trump to blacklist Anthropic for being 'too woke.' As AI technology advances and cybersecurity risks escalate, the public cannot afford to trust Trump’s decisions without oversight, the group said. The complaint highlights concerns that new vulnerabilities discovered in the real world may be due to agencies failing to complete proper reviews or missing important checks. 'We deserve to know what’s in the framework,' the group said, joining calls for greater transparency in AI model deployments. Source: arstechnica

The Trump administration has been rushing to set up a voluntary safety review process since the government flagged Anthropic’s Mythos 5 model as too dangerous to release earlier this summer. The plan was to rapidly expand industry collaborations so that government teams could review prerelease models with reduced or removed safeguards to 'thoroughly evaluate national security-related capabilities and risks.' However, the White House did not identify any companies participating in the GOLD EAGLE clearinghouse or the legal authority for the program. The White House has also refused to share details about the framework, even though it is not classified. Protect Democracy noted that the term 'covered frontier model' is not defined, which could lead to dangerous models slipping through the cracks or agencies being overwhelmed. Source: arstechnica