The White House does not plan to publicly release its new framework for evaluating advanced AI models, three sources familiar with the discussions told Axios1. The administration held staff-level meetings with industry representatives on Tuesday to review the recently completed framework, sharing details with OpenAI, Anthropic, and other AI labs3.
The framework is designed to give the government a process for reviewing frontier models prior to launch2. It covers confidentiality, cybersecurity, insider-risk, intellectual property, and nondisclosure requirements. The framework is expected to set the rules for government access to AI models up to 30 days before release.
The voluntary framework was outlined in a June executive order. That executive order explicitly states that the benchmarking process to assess advanced cyber capabilities of AI models will be classified. There is no requirement in the executive order to publicly release the framework.
The decision to keep the framework private means companies, policymakers, researchers, and U.S. allies outside the process will not have visibility into how the administration plans to implement one of its key AI policies. Companies that were not invited to Tuesday's meetings remain in the dark about the framework's contents.
It is also unclear which "trusted partners" will get early access to advanced models under the framework, including whether any foreign governments would qualify. The European Union declined to comment and the U.K. did not respond to multiple requests for comment. The White House declined to comment.
The meetings included representatives from Anthropic. Nvidia staff also participated. Nvidia CEO Jensen Huang was in Washington, D.C., last week meeting with Commerce Secretary Howard Lutnick and President Trump.
Open-source models were discussed during the meetings, sources familiar with the talks said, without elaborating. The framework arrives as the industry contends with high-profile cyber-attacks and rapid advances by Chinese AI developers, which have reignited debate over how to handle open-source models.
ANALYSIS The secrecy surrounding the framework creates an asymmetry between the AI companies inside the process and those excluded from it — as well as between the U.S. government and allied governments that may seek to coordinate on frontier-model oversight. The classified benchmarking process and the absence of any public-release requirement mean that external scrutiny of the framework's scope and rigor will be limited to whatever participants choose to disclose.