The Trump administration is increasingly dictating access to frontier AI models, shifting significant power away from tech giants like OpenAI and Anthropic. This move comes amid growing concerns over national security and the rapid advancement of AI capabilities globally, particularly from China.
Previously, companies controlled who accessed their most powerful AI tools. However, the White House has launched initiatives like “Gold Eagle” and is reportedly blocking certain AI releases, signaling a new era of government oversight in the AI development landscape.
White House Tightens Grip on AI Frontier Models, Shifting Power from Tech Giants
Sources indicate the Trump administration is increasingly dictating access to cutting-edge AI technologies, a significant shift from the previous model where tech companies controlled distribution.
Key Developments
- The Trump administration is now playing a more active role in controlling the release of advanced AI models, according to sources.
- Previously, AI developers like Anthropic and OpenAI managed access to their frontier models through internal initiatives such as Project Glasswing and Daybreak.
- The White House has launched an AI clearinghouse in partnership with the private sector to address cybersecurity vulnerabilities in AI.
Shifting Control Over AI Access
The Trump administration has implemented new measures to exert greater control over the rollout of future artificial intelligence model releases. The government is now dictating which companies and entities gain access to the latest frontier models, a departure from the previous practice where major American AI firms unilaterally decided on access.
Historically, companies like Anthropic and OpenAI have determined which entities received access to their most powerful AI models, often prioritizing major enterprise clients. Anthropic, for instance, used its Project Glasswing initiative to grant access to its advanced Mythos cybersecurity model to select partners. Similarly, OpenAI, at the request of the administration, has controlled access to its GPT-5.6 release through a consortium known as Daybreak.
While a White House official stated that the administration does not approve AI releases from private companies and that any engagements with government experts are voluntary, recent actions suggest a more direct influence. Last month, the administration reportedly blocked Claude Mythos 5 and Fable 5 citing "national security concerns," only reinstating access after extensive negotiations with Anthropic. In parallel, OpenAI announced it would limit access to new AI models to "trusted partners" to align with government requests.
The Global AI Race and Cybersecurity Concerns
This government intervention comes at a critical juncture, as sophisticated AI tools present significant cybersecurity risks. Concurrently, cheaper, open-weight models, particularly from China, are rapidly narrowing the performance gap with leading American AI labs. For example, Chinese startup Moonshot AI recently unveiled its Kimi K3 model, which rivals the performance of models like Fable and GPT-5.6, even outperforming some U.S. frontier models in independent benchmarks.
David Sacks, founder of Craft Ventures and a former White House AI czar, expressed concern over the Kimi breakthrough, warning that such developments could lead to losing the AI race if other nations do not adhere to the same regulatory constraints.
The administration's approach to AI oversight has been evolving. President Donald Trump's June executive order encouraged companies to voluntarily provide early access to models for government testing. This week, the administration launched "Gold Eagle," a program designed to foster collaboration with the private sector to identify and mitigate cyber vulnerabilities. This initiative positions the White House to greenlight access to new AI models.
The administration's recent moves cast uncertainty on the future of company-led initiatives like Project Glasswing and OpenAI's Daybreak, suggesting that future rollouts will require explicit government approval for partner involvement.
Video Feature
