White House withholds AI testing framework draft from Protect Democracy

The dispute has moved from whether the government must respond to what it will actually hand over. Protect Democracy filed its FOIA request in August and sued on 1 September 2026 against the Office of the National Cyber Director, the Department of Commerce, the Department of Treasury, and the Office of Science and Technology Policy. A September agreement and a 15 September joint status report set a 30 October production deadline, and the government's letter arrived on that same date.

The refusal sits beside a broader push for written rules. Axios reported that OpenAI plans a post-midterms push for mandatory "speed limits" on the most powerful systems. As this piece reads it, the framework's terms decide which models reach customers, and those terms are not public.

Some details have surfaced even without the draft. Schwartz posted that the framework is a seven-page Word document, and that the Office of the National Cyber Director and the Commerce Department appear to be approving "trusted partners" cleared for pre-release access to AI models. The framework's contents, the companies taking part and the legal basis remain undisclosed, according to Protect Democracy. The parties proposed a joint status report for 13 November 2026.