WASHINGTON — The Trump administration has asked OpenAI to limit access to its most powerful model to a small group of government-approved partners, a move that places the White House in an unprecedented role as gatekeeper over frontier artificial intelligence.

The decision, disclosed this week, requires OpenAI to stagger the release of GPT-5.6 rather than rolling it out broadly as the company had planned. Chief Executive Sam Altman told staff Thursday that the rollout would now follow a government-approved schedule, according to people familiar with the discussions.

The request reflects a broader shift in Washington's approach to advanced AI. Rather than waiting for voluntary industry guidelines or relying on congressional legislation, the administration is using direct intervention to shape how the technology is deployed. Three versions of GPT-5.6 will be made available only to partners cleared by U.S. officials, a process one person close to the company described as "ad hoc" vetting.

OpenAI confirmed the changed rollout in a brief statement, citing "government security concerns" as the reason for the restricted access. The model, internally known as GPT-5.6 Sol, is described by the company as a next-generation system with stronger capabilities in coding, science, and cybersecurity.

The arrangement is likely to draw criticism from both libertarian voices worried about executive overreach and AI researchers who argue that limiting access slows the pace of safety-critical testing. It also raises unanswered questions about which officials or agencies will decide who qualifies for access, and under what legal authority.

For now, the policy places the United States inside a small and controversial club of governments that directly control access to advanced AI systems. Its long-term implications for innovation, competition, and the global race for artificial intelligence remain unclear.

Image source: v3b.fal.media