The Pulse
White House Withholds Details of Voluntary AI Testing Plan
The Trump administration declined to release the full text of its voluntary AI model review framework after a FOIA lawsuit by Protect Democracy. A White House letter cited national security and commercial exemptions, leaving the framework’s
AI.info Team · Written with AI · fact-checked ·
“The federal government is asking the public to trust a framework for reviewing AI models before they’re released, but only the government and AI companies know what the framework says. That isn’t acceptable, and we’ll keep fighting in court until the public can see it.”
Vivian Dong, director at the Legal Advocates for Safe Science and Technology
A FOIA deadline ends without the framework
The Trump administration declined to provide the full text of its voluntary AI testing framework by an agreed deadline, according to The Information. In a letter sent October 30, the White House Office of the National Cyber Director invoked national-security and commercial exemptions to withhold the draft and related records sought by the nonprofit Protect Democracy Project.
The letter followed a Freedom of Information Act lawsuit filed by Protect Democracy in September against the National Cyber Director’s office and other federal agencies. The nonprofit had sought information about the framework, the organizations permitted early access to advanced models, and the legal authority for the program. The administration’s response offered some description of the records but did not disclose the framework’s operating details.
Seven pages and a list of trusted partners
The Information reported that the government’s document inventory identifies the “voluntary AI framework” as a seven-page Word document. Most of the other records listed are emails from the National Cyber Director’s office and the Commerce Department to AI labs and vendors, identifying organizations cleared as “trusted partners” for access to frontier models before public release.
The inventory suggests the government is selecting organizations for that early-access role, but the withheld records leave the criteria and names unavailable to the public. The framework’s text is also still undisclosed, so AI companies outside the select group cannot readily assess how the review is supposed to work or what participation entails.
The plan began with a June executive order
President Donald Trump’s June 2 executive order directed federal agencies to develop a voluntary process with AI developers. Under the order, companies could consult the government on whether a model meets the definition of a “covered frontier model,” then provide officials access to such a model for up to 30 days before sharing it with other trusted partners.
The order frames the process around cybersecurity and protection of critical infrastructure. It also says the framework does not authorize mandatory federal licensing, preclearance or permitting of new AI models. That distinction makes the program formally voluntary, while leaving the practical terms of government review and early access dependent on details the administration has not released.
A dispute over public accountability
Protect Democracy’s lawsuit sought records from the White House Office of the National Cyber Director, the Commerce and Treasury departments, and the White House Office of Science and Technology Policy. The group’s case page says its request covered the framework, participating companies and the legal basis for the process. The requested information matters to developers assessing whether to participate, as well as to the public trying to understand how federal officials might influence access to unreleased models.
The administration’s exemptions are its stated grounds for withholding the records; they are not a court ruling that the documents must remain secret. For now, the public can read the executive order’s broad outline, but not the seven-page framework or the records identifying the organizations selected for early access.