Experts, including a prominent think-tank expert, say the US lacks a statutory, predictable process for evaluating frontier AI models, and that a regime focused on closed models concentrates regulatory uncertainty on a few players.

The same critic warns that voluntary review regimes and selective scrutiny imply risk is concentrated among a small set of companies, not evenly distributed across the sector.

At a White House briefing, officials announced a finalized framework but stated they do not plan to release its details publicly.

There is a clear pattern of keeping the framework’s specifics out of the public eye, with no expectation of public disclosure.

The discussion centers on frontier AI models, particularly differentiating between closed and open models in regulatory coverage.

Overall context suggests a balancing act: tighter oversight for closed models to stay competitive with China, while openness for other models complicates comprehensive coverage.

The framework explicitly cannot be used to restrict open models after their release, limiting its reach to non-open deployments.

An executive order issued in June directed major AI developers to submit new models for government review 30 days before release, with a 60-day period to finalize a proposed framework, a deadline that passed without public details.

Major closed-model developers cited include OpenAI, Anthropic and Google, while open-models from Meta and Nvidia are accessible to users, with others like DeepSeek, Moonshot, and Mistral highlighted as benchmarks.

The proposed review targets closed models under tight developer control, exempting open models and several international players from the framework.

Recent autonomous cyber incidents involving AI firms have intensified concerns about AI security and the need for safeguards.

Participants at a White House meeting reportedly included OpenAI, Anthropic, Google, Nvidia, Microsoft and Meta, with discussions framed around a voluntary framework for closed models to maintain US competitiveness against China.