Key facts
- A lawsuit has been filed against four federal agencies by Protect Democracy over the Trump administration's secret AI safety review framework.
- The lawsuit alleges a lack of transparency regarding the framework's text, participants, and criteria for accessing frontier AI models.
- Protect Democracy suggests the framework could be corrupt, potentially favoring AI firms favored by Trump.
- The administration has established a voluntary framework and a clearinghouse called GOLD EAGLE for AI model reviews.
- While some review processes are classified, the framework itself is not designated as classified but details remain undisclosed.
A lawsuit has been filed against four federal agencies by the nonpartisan nonprofit Protect Democracy, demanding the release of information about the Trump administration's secret framework for reviewing advanced AI models. The organization alleges that "almost no details" of this framework have been made public or shared with Congress, leaving it unclear which companies are involved, what criteria are used for reviews, and what legal authority underpins these actions.
Protect Democracy contends that decisions about which AI models are approved and released are critical and that the current process lacks oversight. The lawsuit seeks a court order to compel officials to produce all unclassified procedural and contractual information related to the framework by September 30. The group suggests the framework could be susceptible to corruption, with potential for officials to favor AI firms that Trump supports, while also noting past instances of Trump retaliating against AI companies.
The lawsuit highlights concerns that the executive branch is effectively choosing which companies can release products and which customers gain access to powerful AI technology, without adequate congressional or public scrutiny. This lack of transparency is particularly concerning given the rapid advancement of AI and escalating cybersecurity risks, such as OpenAI's model hacking Hugging Face. The group argues that the public deserves to know the details of the framework to ensure AI is governed safely and in the public interest.
Protect Democracy points to the administration's efforts to establish a voluntary safety review process, including the launch of a clearinghouse called GOLD EAGLE, which relies on industry partners. However, the identities of participating companies and the legal authority for these programs have not been disclosed. While some aspects of the review process are classified, the framework itself is not, yet details remain withheld from the public.