Skip to content
NewsRegulation

White House completes AI model review framework but will not publish it

· by Pondero Newsdesk

The short version

The White House confirmed on August 3 it met the EO 14409 deadline for a voluntary frontier AI review framework, then revealed the next day it has no plans to release the document, leaving the benchmark methodology and coverage threshold both classified.

White House completes AI model review framework but will not publish it

The White House confirmed on August 3 that it completed its voluntary frontier AI review framework by the deadline in President Trump's June 2 executive order, then disclosed the next day that it has no plans to release the document publicly. The result is a governance structure with no published entry threshold, classified benchmarks, and operational details restricted to companies invited to a single industry meeting.

What

The framework stems from Executive Order 14409 ("Promoting Artificial Intelligence Innovation and Security"), signed June 2. It directs an NSA-led interagency group to develop a process giving the federal government up to 30 days of pre-release access to frontier AI models for security evaluation. A White House official told Axios on August 3: "The voluntary framework outlined in the June 2nd executive order was complete by the deadline. Discussions with industry about next steps are underway."

On August 4, the administration held staff-level review meetings with Anthropic, OpenAI, Google, and Meta, per SiliconAngle. Nvidia staff also attended those sessions, per Axios. The day after, Axios reported, citing three sources familiar with the discussions, that the White House does not plan to publicly release the framework document. Companies that were not invited remain unaware of its contents.

The executive order itself classifies the benchmarking methodology used to assess advanced cyber capabilities of models. The threshold determining which models fall within scope is also classified, per Axios, and will be shared with AI developers only "as appropriate." Participation in the framework remains voluntary.

Why it matters

Teams that build on closed frontier models now operate with a supply chain subject to a review process with no published entry criteria, no disclosed timeline structure, and no defined appeal mechanism. Anthropic's frontier models were shut down for 19 days under a related export control order earlier this summer, per TechPolicy.Press. Customers learned of that disruption without advance notice. The new framework creates a parallel mechanism with the same potential for abrupt access changes.

Neither the EU nor the UK was briefed on the framework's contents, per Axios. Open-weight models appear excluded from the current scope; the framework as described targets closed frontier models from companies large enough to receive a direct White House invitation.

Policy researcher Michelle De Mooy published five transparency questions on August 5 in TechPolicy.Press, covering what makes a model subject to review, how long review can last, who sets the access list, what recourse developers have, and what the public will ever be told. Without answers, she argued, the process "will simply have documented unfettered discretion in the hands of the executive branch" rather than a governance framework.

What to watch next

No AI lab has publicly confirmed or declined participation. Congress's Great American Artificial Intelligence Act of 2026 contains transparency mandates and independent audit requirements that would partially address the accountability gap, per TechPolicy.Press. If that bill advances, it would create a statutory alternative to the current arrangement of voluntary participation under classified rules. Whether any lab opts out openly, or whether the benchmark threshold is eventually published in unclassified form, are the two near-term signals to monitor.

Sources