The White House withholds all 60 pages on its voluntary AI framework, and most of the 28 letters on its withholding list confirm which trusted partners may use new models before general release

(letter dated October 30)

Concept illustration: on a desk at left sits an open envelope with a sheet lettered "INDEX" rising out of it, showing three grey lines each followed by a thick black bar, and a person with a hair bun and a lanyard badge, standing to the right of the desk, looks down and holds the sheet with both hands. At right, against a pink background, stands a dark-framed glass-door cabinet under a sign lettered "WHITE HOUSE ONCD", and on its middle shelf is a stack of red folders lettered "AI FRAMEWORK", tied with cream string in a bow. Printed on the image: 60 pages AI framework records, White House withheld all.

The White House Office of the National Cyber Director (ONCD) withheld all 60 pages of records on its voluntary AI framework on October 30, in a Freedom of Information Act lawsuit. Our inference: AI labs have sent at least some of their early-access customer lists to the government before a new model's launch, so a company joining a lab's preview should assume its name may reach the White House.

The withholding list the government is required to attach names 28 letters sent to AI labs, one by one. In a records lawsuit, this list is the catalog the government must file when it holds documents back, giving each record's date, type and a one-line description. Of the 28 letters, 27 came from ONCD and 1 from the Commerce Department. Most descriptions read "identifying trusted partners cleared for pre-release access to an identified covered model", that is, confirming which trusted partners may use a new model before it opens to everyone (ONCD response letter and withholding list, 2026-10-30). 'Covered model' is the framework's own term. A few say only that they identify trusted partners. Of the rest, one concerns a proposed list and several concern testing and development partners. Lab and model names are all redacted as "[AI Lab]" and "[Covered Model]".

The government's stated reason for withholding says the labs supplied these lists. They are "derived from confidential customer lists submitted to ONCD". When OpenAI released GPT-5.6 on June 26, OpenAI's announcement said that, at the government's request, it was starting with a limited preview for a small group of trusted partners "whose participation has been shared with the government". The Commerce Department letter on the list is also dated June 26. Its lab name is redacted, so the list doesn't show whether that letter went to OpenAI.

The clearance letters began before the framework file's date. Among the 60 withheld pages, the framework is a 7-page Word file titled "Voluntary AI Framework for Covered Models", dated July 31. The earliest letter on the list is dated June 25, 36 days before that file date, and 15 more letters went out between July 31 and September 17. A file date isn't necessarily the day the rules were first written down. The same list also includes a May 24 draft of an AI safety executive order, which the government treats as a source document for the framework and withholds along with it. Axios, citing anonymous sources, reported that the framework is a set of voluntary rules the White House showed AI companies in early August, covering only closed frontier models.

The plaintiff narrowed the request before this response came back. Protect Democracy Project, the US nonprofit legal advocacy group that brought this lawsuit, and the government filed a joint report with the court in September that narrowed the request to a few categories, including the framework text and the trusted-partner lists. In its October 30 response letter, the government says this is its complete response under that report. It cites four exemptions: national security, the Cybersecurity Information Sharing Act, trade secrets, and presidential communications and internal deliberations, and releases not a single page. Leo Schwartz, a reporter at The Information, pointed to the withholding list on X on November 1. Miles Brundage, OpenAI's former head of policy research, criticized on July 17 a licensing regime that is voluntary in name but required in practice, on the grounds that "the rules don't have to be written down like actual regulations". Brundage didn't say which document the criticism referred to.

Not proven yet: the list doesn't show whether the government ever rejected a customer How we checked →

Next to watch: whether the plaintiff challenges the withholding list in court in the coming weeks, and whether the court treats the customer lists labs submitted as trade secrets. If another frontier lab says its early-access lists never went to the government, the practice may belong to one lab only.

Source and checks

This item ran in the brief of Nov 2, 2026, alongside 4 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.

Cite this item

SecondSource (2026). "The White House withholds all 60 pages on its voluntary AI framework, and most of the 28 letters on its withholding list confirm which trusted partners may use new models before general release." https://secondsource.io/en/news/2026-11-02-1/. Accessed access date.