White House Withholds AI Review Framework Records

Office of the National Cyber Director withheld all 60 pages of records sought under FOIA, citing national security and confidential commercial information exemptions.

Updated on Nov. 3, 2026 in Artificial Intelligence

White House Withholds AI Review Framework Records

Live Poll

Do you trust the federal government to review powerful AI models without public disclosure?

The White House withheld its entire 60-page frontier AI review framework records in response to a Freedom of Information Act request. The refusal arrived on the Oct. 30 deadline in an open records lawsuit filed by Protect Democracy in U.S. District Court for the District of Columbia.

Why it matters

The framework governs a voluntary process under which developers can give the government access to covered frontier AI models for up to 30 days before release. Withholding the document in full limits public scrutiny of how models are selected for review in the United States.

The National Security Agency uses classified benchmarks for advanced cyber capabilities to determine which models qualify for review, and those benchmarks are not public. The government committed to producing a Vaughn index describing the withheld records by Oct. 30, but the underlying criteria remain undisclosed.

The players

Protect Democracy

A nonpartisan nonprofit organization that litigates government transparency and accountability cases, and filed the open records lawsuit over the framework.

Office of the National Cyber Director

A White House office responsible for coordinating national cybersecurity policy, which withheld the framework records under FOIA exemptions.

National Security Agency

The U.S. intelligence agency that bears primary responsibility for deciding which frontier AI models qualify for the voluntary government review.

The details

President Donald Trump created the review process through a June 2 executive order that explicitly rules out mandatory licensing or preclearance. The White House says it completed the framework by the Aug. 1 deadline and discussed it with OpenAI, Anthropic, Google, Meta and other developers.

June 2, 2026: Trump signed the executive order creating the voluntary AI review process.
Aug. 1, 2026: The White House says the framework was completed by this deadline.
Oct. 23, 2026: The government told a court it had found the framework.
Oct. 30, 2026: The deadline for release of non-exempt records; the withholding letter was issued.

Roadmap

The June 2, 2026 executive order created a voluntary, non-licensing model for overseeing frontier AI, relying on the National Security Agency to classify which models qualify. The withholding of the implementing framework means the public still cannot see how that voluntary model is being applied.

The framework determines which AI models get a pre-release government look, which could shape how quickly new AI tools reach consumers and businesses. Because the criteria are withheld, users and developers cannot yet tell which systems will face that review.

The takeaway

Withheld government records limit public understanding of how AI oversight works in practice. Following the outcome of the Protect Democracy lawsuit and the forthcoming Vaughn index will show how much of the framework becomes public.