The White House has asked OpenAI to gate the release of GPT-5.6 to a small set of government-approved customers, with federal officials signing off one client at a time. According to people familiar with the matter cited by CNN, Axios, and The Information, it’s the first time Washington has preemptively bottled up a frontier model before any public launch.

The request came from the Office of the National Cyber Director and the Office of Science and Technology Policy. Sam Altman, chief executive of OpenAI, discussed GPT-5.6 on Wednesday with Commerce Secretary Howard Lutnick, who sought assurance every relevant agency had tested and approved the system.

In an internal Q&A the same day, Altman told staff the company would proceed with a limited preview, framing it as the “fastest path to a broad release” that would follow “a couple of weeks later.” In a memo obtained by The Information, he also said OpenAI had “made clear to the U.S. government that this is not our preferred long term model.” Two audiences, two registers, one product.

The backdrop is sharper than the announcement suggests. Earlier this month the Commerce Department forced Anthropic to pull its Mythos and Fable 5 models offline over cybersecurity concerns, and officials reportedly view GPT-5.6 as “on par” with Mythos. A June 2 executive order directs voluntary cooperation between frontier labs and federal agencies on security review, though no formal framework yet exists.

What’s emerging looks less like the export-control regime of the 2022 chip rules and more like the pre-clearance posture once reserved for cryptographic exports in the 1990s, applied this time to inference itself. OpenAI declined to comment.

Sources