The Trump administration has taken a more direct role in deciding which companies can access the most advanced AI models from Anthropic and OpenAI, according to CNBC, which cited two people familiar with the matter. The shift moves a decision labs previously made largely on their own into the hands of the government.
Until recently, each lab ran its own vetting. Anthropic’s Project Glasswing decided which partners could use its Claude Mythos cybersecurity model, and OpenAI’s Daybreak consortium played a similar role for cyber-focused releases; OpenAI was also asked to gate its recent GPT-5.6 launch and limit early access to “trusted partners,” CNBC reported.
A new clearinghouse
That vetting has now folded into Gold Eagle, a White House-led clearinghouse launched July 14 to speed up the discovery and patching of software vulnerabilities using AI, according to the official announcement. It is run by the Treasury Department alongside the Department of Homeland Security’s Cybersecurity and Infrastructure Security Agency and the Department of War, working with what the announcement calls “America’s most critical and innovative private sector companies.” One source told CNBC the arrangement effectively puts “the White House in charge of greenlighting which companies can access new AI models.”
A White House official disputed that framing, telling CNBC that “decisions on timing and scope of releases rest entirely with the companies” and that engagement with government reviewers remains “voluntary.” The administration, the official added, “continues to collaborate with all of America’s frontier labs to strengthen the security of this technology without stifling innovation.”
A rockier rollout pattern
The reported shift follows a bumpy few weeks for model access. Last month the administration briefly blocked Claude Mythos 5 and Fable 5 over national-security concerns, restoring access only after Anthropic negotiated directly with officials. Trump’s June executive order had asked AI companies to voluntarily submit new models for early government testing — a regulation framework CNBC’s sources say has since expanded well past that original scope.
The pressure to keep releases moving quickly has grown since Moonshot AI’s Kimi K3 matched the performance of leading US systems at a fraction of the cost, a milestone that DeepMind CEO Demis Hassabis has cited in his own push for clearer, faster AI oversight. Former White House AI advisor David Sacks argued this week that heavier gatekeeping risks costing the US its lead in the field.