Lawsuit seeks to force disclosure of Trump administration's secret AI safety review framework
Nonprofits and state officials are challenging the Trump administration's classified AI model review process, alleging lack of transparency and potential corruption in decisions about which AI companies can release products.
What happened
Protect Democracy, a nonpartisan nonprofit, filed a lawsuit against four federal agencies (Office of the National Cyber Director, Office of Science and Technology Policy, Treasury Department, and Commerce Department) on Wednesday demanding disclosure of the framework used to conduct safety reviews of frontier AI models. The complaint alleges that almost no details have been made public about the review process, which companies participate, what legal authority officials have, or how access decisions are made. The lawsuit seeks production of unclassified procedural and contractual documents by September 30, including the framework's text, participation terms, participant identities, and criteria for granting or withholding access. The White House announced on August 3 that it completed a voluntary framework for reviewing AI models and is actively using it. In July, the White House launched GOLD EAGLE, a cybersecurity clearinghouse that similarly lacks public disclosure of participating companies.
Context
The Trump administration's secretive approach to AI model review concentrates decisions about technology that could shape American industry, national security, and global economies entirely within the executive branch with no congressional or public oversight. Protect Democracy argues this creates risk of corruption, given Trump's history of retaliating against companies he deems insufficiently aligned with his preferences—citing a recent judicial ruling that Trump's attempted blacklist of Anthropic was illegal. The group contends the public cannot verify whether vulnerabilities discovered in deployed AI models result from inadequate review processes, outdated testing, or missed safety checks by expert standards. This stands in contrast to California's proposed SB 813, which would establish independent-led safety standards developed transparently with public input and accountability. Congressman Greg Casar and California State Senator Josh Becker have criticized the administration's approach as failing to protect national security and public safety while benefiting AI companies.