Protect Democracy, formally the Protect Democracy Project, is a United States nonpartisan nonprofit organization whose stated mission is to "prevent American democracy from declining into a more authoritarian form of government." It works substantially through litigation and Freedom of Information Act requests. In 2026 it became the principal litigant seeking disclosure of the executive branch's framework for reviewing frontier AI models before public release — the process described at AI Pre-Release Vetting — making it the main external source of information about a program the administration has described only in general terms.
FOIA litigation over the frontier-model review framework
In August 2026 Protect Democracy submitted identical FOIA requests to four agencies — the White House Office of the National Cyber Director, the White House Office of Science and Technology Policy, the Department of Commerce and the Department of the Treasury — seeking non-classified information about the "voluntary" framework the White House said on August 3, 2026 it had finalized for reviewing advanced AI models before release, the identity of participating companies, and the legal authority for the framework. The organization requested expedited processing. Only the Office of the National Cyber Director responded, and it denied expedition (Source: arstechnica.com).
On September 1, 2026 Protect Democracy, represented by Kaplan Johnson Abate & Bird LLP, filed suit against the four agencies to enforce the requests, in an action captioned Protect Democracy Project v. Office of the National Cyber Director, No. 1:26-cv-03064, in the United States District Court for the District of Columbia before Judge Ana C. Reyes — tracked at Protect Democracy Project v. Office of the National Cyber Director (Source: courtlistener.com; protectdemocracy.org; cdn.arstechnica.net). The complaint seeks an order requiring production by September 30 of what it calls the "unclassified procedural and contractual architecture: the framework's text, the terms of participation, the identity of participants, and the process and criteria by which access to frontier models is granted or withheld," together with an injunction against improper withholding of unclassified records.
The complaint's stated grounds are that the framework is not designated classified; that the term "covered frontier model" is undefined; that the White House has not identified any company participating in the GOLD EAGLE program, the terms of participation, or any legal authority for it; and that the framework has been shared only with "a select set of AI companies," excluding smaller developers and AI safety organizations. Protect Democracy also argues that Congress needs the records to decide whether to extend the Cybersecurity Information Sharing Act liability protections that, on the administration's own account, sustain GOLD EAGLE. Deana El-Mallawany, the organization's director of impact programs and counsel, said that neither Protect Democracy nor Congress has enough information about the program to make an informed decision on that extension, and that any documents obtained would be posted publicly (Source: arstechnica.com).
El-Mallawany has set out the risks the organization attributes to an undisclosed review process: that the White House could use it to press AI companies over deployment decisions such as lethal autonomous weapons and mass surveillance, to reward or punish companies for political alignment, or to require particular viewpoints in models — and, separately, that the framework may simply be ineffective and create a false sense of security (Source: arstechnica.com). The organization's own case page states that the regulatory process "was set up without any legislation or legal authority of any kind" (Source: protectdemocracy.org).
Protect Democracy records an initial production of uncovered records dated September 8, 2026 (Source: protectdemocracy.org). Gary Marcus published an account on September 15, 2026 describing a production of 132 pages in which nearly all substantive content is redacted, and stating that the records confirm participation in the discussions by OSTP Director Michael Kratsios and US Chief Technology Officer Ethan Klein; the same account reports El-Mallawany as saying the organization expects to agree a further production schedule (Source: garymarcus.substack.com).
Open questions
- Marcus's account attributes the 132-page production to OSTP, while the complaint and the organization's own case page identify four defendant agencies and record ONCD as the only agency that responded to the expedition request. Which agency produced the pages is not resolved by the sources held.
- Whether the September 30 production date sought in the complaint was ordered, met, or superseded is not recorded. The docket shows the preliminary-injunction motion that would have produced a ruling on it was withdrawn on September 15, 2026, and the schedule the parties agreed instead is not public (Source: courtlistener.com).
Relationships
- litigates: Protect Democracy Project v. Office of the National Cyber Director — against Office of Science and Technology Policy (OSTP) and three other agencies
- depends-on: AI Pre-Release Vetting (the program its litigation is about)
- related: Gary Marcus (published the September 2026 production); AI Federalism; Frontier AI Governance