Emerging Technology — 2026-08-03

White House Finalizes Classified AI Framework Behind Closed Doors but Refuses to Disclose Contents

BLUFBy classifying benchmarks and coverage thresholds while keeping the framework nominally voluntary, the administration retains unilateral control over which AI models face scrutiny, making substantive disclosure by September 2 unlikely.

A White House official said the voluntary AI evaluation framework directed by the June 2 executive order was completed by the deadline but declined to disclose its contents, who has reviewed it, or when it takes effect 12. AI companies including Anthropic, Google, Meta and OpenAI are expected to review the framework Tuesday at a meeting with the Office of the National Cyber Director, according to Politico, which cited five people familiar with the planning 2. Reuters reported the finalized framework includes voluntary cybersecurity tests to measure hacking capabilities of the most advanced US AI models, though the White House official did not detail how results would be reported or whether they would be made public 3. Axios reported the executive order classifies the benchmarking process for assessing advanced cyber capabilities and the threshold determining which models are covered, while the framework itself was not designated classified 1.

Analysis
The administration's refusal to disclose the framework's contents, reviewers, or effective date signals intent to control the compliance narrative before industry can contest thresholds set outside public view, and substantive disclosure by September 2 is unlikely given that the executive order already classifies the benchmarking methodology and coverage threshold, leaving officials no incentive to reveal more. This judgment carries high confidence, based on consistent official statements across Reuters, Politico, and Axios that the document exists but remains withheld, with independent primary-sourced reporting from each producing broad corroboration rather than single-wire repetition. The opacity may instead reflect unfinished internal consensus on scope and reporting mechanics rather than deliberate concealment. Continued silence leaves smaller AI developers outside the four major labs unable to calibrate whether their models trigger the 30-day pre-release government access requirement, forcing them to plan compliance blind.

3 sources

White House finalizes AI framework behind closed doors - Axios
White House finalizes voluntary AI oversight framework - Politico
US finalizes voluntary AI safety tests, White House official says - Reuters