Analysis
The White House completed its voluntary framework for reviewing advanced AI models before release, hitting the August 1 deadline set by a June executive order -- and then confirmed it will not make the document public, name which companies or models have been reviewed, or disclose when the framework actually takes effect, according to [Fortune](https://fortune.com/2026/08/04/baffling-white-house-wont-publicly-release-ai-model-evaluation-framework-it-reviewed-today-with-openai-anthropic-microsoft-and-others/).
Who Was in the Room
Meta, Nvidia, Microsoft, OpenAI and Anthropic all attended an August 4 meeting to review the completed framework alongside several smaller AI companies, giving the largest labs direct visibility into a process the general public won't see.
What Stays Classified
Under the framework, covered frontier labs must submit qualifying models to government review up to 30 days before public release -- but the benchmark and threshold defining which models actually qualify for that review are themselves classified, with the executive order allowing those specifics to be shared with AI developers and researchers only as officials see fit. That's a real departure from how the EU's AI Act and most prior US technology regulation have worked, trading public accountability and outside scrutiny for what the administration frames as necessary national-security discretion.
What to watch: whether any of the labs that attended the August 4 meeting push for greater public disclosure once they've actually started operating under the framework's submission requirements, and whether Congress or outside researchers gain any formal channel to review classified review outcomes over time.
