GPT-5.6 Launch Staggered at the White House’s Request

OpenAI is staggering the public release of GPT-5.6 after the Trump administration asked the company to limit who can use it first, according to a report from The Information on June 25. Rather than the broad, all-at-once launch the company has favored for past models, GPT-5.6 will go initially to a narrow set of government-approved customers — with Washington signing off on access, the report says, customer by customer.

CEO Sam Altman laid out the arrangement to employees during an internal Q&A the same day. GPT-5.6 will first reach only a select group of enterprise partners, with the government “approving access customer by customer during this preview period,” before a broader rollout expected “a couple of weeks” later — and only if the government-managed approval process keeps pace. Altman was candid that the setup is a concession, telling staff it is “not our preferred long-term model, and will work with them and others in industry to achieve a more sustainable approach.”

The request did not come from a single office. The Information reports that the Office of the National Cyber Director, the Office of Science and Technology Policy, and Commerce Secretary Howard Lutnick all shaped the ask, which was framed around national-security concerns about the model's advanced capabilities. It is an unusual degree of government involvement in the launch of a commercial product — and a sign that frontier AI releases are increasingly being treated as matters of state, not just shipping decisions.

The move also explains a swing that AI-watchers had already noticed. On the prediction market Polymarket, the odds of a GPT-5.6 launch during the June 22–28 window collapsed from roughly 83% earlier in the month to about 18%, with more than $560,000 wagered on the timing. Traders now price a release by the end of July at around 94%. What looked for weeks like a near-certain late-June arrival has quietly slipped — and the gating is why. (For the full rundown of what GPT-5.6 is rumored to include, see our explainer on everything we know so far.)

The episode rhymes with the one that engulfed Anthropic earlier this month, when the U.S. government ordered the company to pull Claude Fable 5 and Mythos 5 offline worldwide under an export-control directive. Different mechanism — a takedown there, a staggered launch here — but the same underlying pattern: as the most capable models cross new capability thresholds, Washington is moving to control who gets to use them, and when. For OpenAI, the immediate cost is a slower, narrower debut for a model it had hoped to put in front of hundreds of millions of users.