Protect Democracy sues federal agencies seeking release of secret AI safety rules
The lawsuit demands the unclassified framework of the Trump administration's GOLD EAGLE program, alleging private agreements with OpenAI and political blacklists.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- Nonprofit group Protect Democracy filed a lawsuit against four federal agencies demanding the public release of the GOLD EAGLE AI safety review framework, according to Ars Technica.
- The lawsuit alleges the Trump administration's voluntary framework is used to secretly regulate the release of frontier models, citing a private distribution agreement with OpenAI.
- Legal experts argue the executive branch is using the Cybersecurity Information Sharing Act (CISA) of 2015 as authority, creating severe risks of political coercion and extortion of AI firms.
Nonpartisan nonprofit group Protect Democracy has filed a federal lawsuit against four federal agencies — including the Office of the National Cyber Director and the Commerce Department — seeking the public release of the secret framework used by the Trump administration to vet frontier AI models, Ars Technica reported on September 2. The legal challenge targets the 'GOLD EAGLE' program, a voluntary clearinghouse launched in July 2026 to review AI models before public release.
The complaint alleges that the administration has operated the program in near-total secrecy, leaving Congress, tech experts, and the public with no oversight. According to the lawsuit, the executive branch is using the program to secretly control which companies can release AI products and which customers can buy them. The suit alleges OpenAI negotiated a private agreement under the program to restrict model distribution to 'government-vetted partners,' and points to the illegal blacklisting of Anthropic and the administrative hold on Anthropic's Mythos 5 model as evidence of potential abuse.
Legal analysts believe the administration is invoking the Cybersecurity Information Sharing Act (CISA) of 2015 as the statutory authority for GOLD EAGLE's liability protections. Protect Democracy's lawsuit seeks a court order forcing the agencies to produce unclassified details of the framework, its participants, and its criteria by September 30, 2026. Protect Democracy argues that the secret reviews create abundant risks of political coercion, where the White House could extort AI startups or force them to deploy models for lethal military use or mass surveillance under threat of withholding release approval.
When a voluntary, unclassified AI safety framework is operated in absolute secrecy, it ceases to be a safety gate and becomes a mechanism for executive leverage and backroom deals. Protect Democracy's lawsuit exposes the danger of using decade-old cybersecurity legislation to secretly regulate frontier models without congressional oversight. Forcing transparency on the GOLD EAGLE program is critical to preventing the federal government from using safety reviews as a political weapon to favor specific corporate partners or punish perceived opposition.