Trump Administration Refuses to Release Information on AI Testing Framework

The Trump administration declined to produce key documents related to its voluntary AI testing framework, despite an agreement struck in September with the nonprofit Protect Democracy Project after a Freedom of Information Act lawsuit. In a letter sent to the nonprofit on Friday by the White House’s Office of the National Cyber Director and shared with The Information, the Trump administration’s FOIA office cited national security and commercial exemptions to argue that it did not have to share a full draft of the framework, as well as related documents including records of companies and other organizations allowed early access to frontier AI models. The letter did provide some details on the White House’s “trusted partner” program, as well as the scope of the framework.

The framework, which was created by a June executive order, has been a source of consternation for AI companies producing models or using other labs’ models. It lays out a process for frontier AI companies such as Anthropic and OpenAI to submit their new models for government review before releasing them to the public. Still, after finalizing the framework in early August, the White House has only briefed a select group of companies and refused to share details more broadly on how the process works. That has left unanswered questions for many companies on key aspects of the framework, including which companies qualify as “trusted partners” and can receive advanced access to frontier models before their wide release.

The Office of the National Cyber Director’s letter sent on Friday does provide some answers. In a list of documents that it produced in response to the FOIA lawsuit, the White House details a “voluntary AI framework” consisting of a seven-page word document. The majority of the rest of the documents are emails from the Office of the National Cyber Director and the Commerce Department to AI labs and vendors identifying trusted partners cleared for pre-release access to frontier models, signaling that the government is approving a “white list” of organizations for the trusted partner program.

“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,” said Vivian Dong, a director at the Protect Democracy project, in a statement. “That isn’t acceptable, and we’ll keep fighting in court until the public can see it.”

The White House did not respond to a request for comment.