President Donald Trump points while signing an executive order on AI next to Senate Commerce Committee Chairman Ted Cruz (R-Texas) and Commerce Secretary Howard Lutnick in the Oval Office at the White House on December 11, 2025 in Washington.
Al Drago | Reuters
The Trump administration has taken new steps to gain greater control over the rollout of future artificial intelligence model releases by determining which companies and entities are granted access to the latest frontier models, two sources familiar with the matter told CNBC.
Until now, that decision has been in the hands of America’s AI giants.
both human and OpenAI They decide which companies and agencies have access to their most powerful models, and often include large enterprise customers.
Anthropic has announced its most capable Mythos cybersecurity model to a handful of Project Glasswing partners. OpenAI has been asked by governments to gate the recent GPT-5.6 release and has a similar consortium called Daybreak when it comes to cybersecurity models.
White House officials told CNBC that they have not given approval for AI releases from private companies.
The official told CNBC, referring to President Trump’s recent executive order, that any engagement, testing or meeting with government experts is “voluntary” and “decisions regarding the timing and scope of releases are left entirely to the companies.”
“The government continues to work with all of the nation’s frontier laboratories to strengthen the security of this technology without stifling innovation,” they wrote.
But the Trump administration blocked Claude Mythos 5 and Fable 5 last month, citing “national security concerns,” and restored access after weeks of intense negotiations with Anthropic. OpenAI announced last month that it would limit new AI models to “trusted partners” to comply with government mandates.
The White House is walking a fine line when it comes to regulation, as advanced AI tools pose significant cybersecurity risks and cheap open-weight models from China are rapidly closing the gap with America’s Frontier Labs.
Chinese startup Moonshot AI on Friday announced its Kimi K3 model, which has nearly caught up to the performance of Fable and GPT-5.6, and even outperformed the US Frontier model in at least one independent benchmark.
David Sachs, founder of Craft Ventures and former White House AI czar, called Kimi’s rise “concerning.”
“This is how we lose the AI race,” he wrote. “If we get stuck, other countries won’t follow our rules either.”
The government has already taken several steps in recent months to reshape its oversight of AI, starting with President Donald Trump’s June executive order requiring companies to voluntarily provide the government with early access to testing models.
This week, the administration launched its own program called Gold Eagle, which aims to partner with the private sector to discover and fix cyber vulnerabilities.
The so-called clearinghouse will be responsible for the White House authorizing which companies have access to new AI models, said the people, who asked not to be identified to discuss private information.
The administration’s move leaves the future of corporate-led initiatives like Project Glasswing and OpenAI’s Daybreak in doubt.
One source said that from now on, these developments will require explicit government approval, allowing partners to be involved.
— CNBC’s Megan Cassella and Ashley Caputo contributed reporting
