White House tightens control over access to frontier AI models

The White House is taking a more active role in deciding which organisations can access the latest frontier artificial intelligence models, marking a shift from the previous industry-led approach, according to a CNBC report citing people familiar with the matter.

The move comes as the Trump administration seeks to strengthen oversight of advanced AI systems while addressing growing national security and cybersecurity concerns.

AI companies previously controlled model access

Until recently, companies including OpenAI and Anthropic decided which businesses, researchers and government agencies could use their newest AI models.

Anthropic reportedly limited access to its Mythos cybersecurity model through Project Glasswing, while OpenAI managed access to its latest GPT-5.6 model under a similar initiative called Daybreak.

However, a White House official told CNBC that the government does not approve AI model releases by private companies, describing any testing or collaboration with federal experts as voluntary.

"The Administration continues to collaborate with all of America's frontier labs to strengthen the security of this technology without stifling innovation," the official said.

Administration expands AI oversight

According to CNBC, the administration recently launched a cybersecurity initiative known as "Gold Eagle," designed to work with private companies to identify and address vulnerabilities in advanced AI systems.

The programme reportedly includes a government-led clearinghouse that could determine which organisations receive early access to new AI models.

The report also said future partner rollouts could require explicit government approval, although the White House disputed suggestions that it authorises private AI releases.

National security concerns drive policy

The reported changes follow heightened concerns over the security risks posed by increasingly powerful AI models.

Last month, Anthropic's Claude Mythos 5 and Fable 5 models were temporarily blocked over national security concerns before access was restored following discussions with the government, according to the report.

OpenAI also said it would restrict access to certain new AI models to trusted partners in response to government requests.

Competition with China intensifies

The policy shift comes as Chinese AI firms continue to narrow the performance gap with U.S. developers.

On Friday, Chinese startup Moonshot AI unveiled its Kimi K3 model, which reportedly matched or exceeded leading American models in some independent benchmarks.

Former White House AI adviser David Sacks described the development as "concerning," warning that excessive restrictions could weaken the United States' position in the global AI race.