The White House has finalized a voluntary framework for reviewing the cybersecurity capabilities of the most advanced artificial intelligence models before they are released, marking a new step in the United States' approach to artificial intelligence (AI) oversight.

White House officials met with representatives from OpenAI, Anthropic, Google, Meta and Nvidia on Tuesday to review the framework, which focuses on evaluating the cybersecurity capabilities of advanced AI models.

Under the framework, developers of the most advanced AI models will be asked to provide the US government with access to their systems for up to 30 days before public release. The review is intended to assess the advanced cybersecurity capabilities of AI models and the potential risks they may pose.

The framework does not cover open-weight AI models, whose model weights are publicly available. Instead, it focuses on the most advanced closed models developed by companies such as OpenAI, Google and Anthropic. The distinction has drawn attention as policymakers and industry continue to debate how advanced AI should be reviewed.

Participation in the program is voluntary, and the framework itself has not been made public. According to US media reports, only a limited number of companies were invited to review the details during White House meetings.

The administration has said little publicly about how the reviews will be conducted or how models will be evaluated. Some industry observers and policy groups have questioned whether a voluntary and non-public framework will provide sufficient transparency and oversight.