The Trump administration privately briefed a small number of technology companies on Tuesday about a new voluntary framework for government reviews of advanced AI models, but it will not make the framework public, people familiar with the meeting told Semafor. Axios reported the development earlier.
Sources at the closed-door meeting said the decision not to publish the framework has stirred concern in industry circles: companies that were not briefed may choose not to participate in government reviews because they lack sufficient information about the process and requirements.
Ironically, administration officials also told the invited firms that they plan to host a public event later this month to promote the framework.
Adam Thierer of the R Street Institute told Semafor: “The Trump team appears destined to give us something far more arbitrary and burdensome [than the Biden administration] with this behind-closed-doors de facto licensing regime they are concocting.”
The White House did not respond to requests for comment about why it is keeping the framework private or about the upcoming public event.
The briefing, which provided invited companies with details of the framework, took place amid rising pressure on the White House to address Chinese developments in artificial intelligence. Hours after the meeting ended, OpenAI reported two additional security breaches.
Why this matters
- Keeping the framework private raises transparency concerns: firms that were not invited have less information to decide whether to opt into a voluntary government review.
- The contrast between secret briefings and a planned public promotion may deepen mistrust between industry and the government.
- The timing — with heightened focus on Chinese AI and new security incidents at OpenAI — highlights the urgency and complexity of designing oversight for advanced AI systems.
Based on available reporting, it is not yet public which specific companies were invited to the briefing, nor are the detailed requirements or procedures of the administration’s framework known.
