For years, companies like Anthropic and OpenAI decided who got to use their most powerful AI models. That has changed.
The White House recently began exercising greater authority over who can utilize the most advanced frontier models. Power shifted quietly over the past several months, and it became official this week.
Last month, the White House blocked the release of two Anthropic models called Claude Mythos 5 and Fable 5, citing national security concerns. That was the first time the government stepped in before a model even hit the market. After weeks of intense negotiations, the administration allowed them out, but the message was clear.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
In June, the president issued an executive order that requested firms to provide the government with advance access to their AI models for evaluation. Just this week, the government initiated the Gold Eagle program, which works with industry partners to identify and resolve cybersecurity weaknesses.
A White House official, speaking on condition of anonymity, stated that the administration does not grant permissions for AI product launches by private firms. According to the official, any interactions, evaluations, or discussions with government specialists are "voluntary," and "decisions on timing and scope of releases rest entirely with the companies." The official also said "the Administration continues to collaborate with all of America's frontier labs to strengthen the security of this technology without stifling innovation."
Sacks, who previously served as White House AI czar, described the Kimi development as "concerning." He wrote, "This is how you lose the AI race. The rest of the world won't play by our rules if we bog ourselves down."
The administration is carefully balancing regulation amid a period when advanced AI technologies present significant cyber threats, and lower-cost, open-weight models from China are rapidly narrowing the performance gap with U.S. cutting-edge systems. This tension between security and competitiveness is not new - similar debates have played out over encryption and semiconductor export controls. But here the stakes are higher because AI models can be easily copied and shared, making unilateral controls difficult to enforce.
Anthropic had been running Project Glasswing, using its Mythos cybersecurity model with a handful of partners. The administration requested that OpenAI restrict access to its latest GPT-5.6 version, and the company also operates a comparable group named Daybreak for its security-focused AI. Both companies previously decided which partners joined those programs.
According to an individual familiar with the matter, future deployments will need specific authorization from the government regarding eligible partners. These government actions have cast uncertainty on the future of industry-led efforts like Anthropic's Project Glasswing and OpenAI's Daybreak.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets
