The White House does not intend to make public the voluntary framework it has developed for evaluating advanced artificial intelligence models, three sources familiar with the discussions told Axios.
Why this matters
The framework could have global implications for AI security, yet its details will be disclosed only to the companies participating in the process. Keeping the document private leaves other companies, policymakers, researchers and some U.S. allies without clarity on how the administration plans to implement one of its central AI policies.
Meetings between government and industry
On Tuesday, the White House held staff-level meetings with industry representatives to review the recently completed framework. Firms that were not invited to those sessions remain unaware of the framework's contents.
Sources said open-source models were discussed at the meetings, though they did not provide specifics. They also said Nvidia staff participated in the sessions.
Security and geopolitical context
The framework comes as the industry faces high-profile cyber-attacks and rapid advances by Chinese AI developers, renewing debate over how to handle open-source models. It is unclear which "trusted partners" will receive early access to advanced models under the framework and whether any foreign governments would qualify.
The European Union declined to comment, and the United Kingdom did not respond to multiple requests. The White House also declined to comment.
Relevant provisions from the executive order
The voluntary framework is outlined in a June executive order and is meant to provide a process for AI developers to work with the government to determine whether models under development fall within the order's scope.
The framework is expected to set terms for government access to AI models up to 30 days before public release, addressing confidentiality, cybersecurity, insider-risk, intellectual property and nondisclosure requirements.
The executive order explicitly states that the benchmarking process to assess advanced cyber capabilities of AI models will be classified. The order does not require the voluntary framework to be released publicly.
