OpenAI has agreed to roll out its next major model, GPT-5.6, in stages — starting with a small set of government-vetted enterprise customers — after a request from the Trump administration to limit its release on cybersecurity grounds, according to multiple reports.

What is reported

The arrangement was first reported by The Information and corroborated by Axios, Engadget and others, based on an internal memo and people familiar with the matter. OpenAI chief executive Sam Altman is said to have told staff that the federal government would approve access "customer by customer" during a preview period, with a broader release expected weeks later if the review goes smoothly. Reports say the model will initially reach roughly 20 partners, with access routed through a major cloud platform.

OpenAI had not issued an on-the-record public statement at the time of writing, and several specifics rest on anonymous sourcing.

What the administration asked, and why

According to the reporting, the request came from White House offices dealing with cybersecurity and science-and-technology policy, with the stated concern that powerful AI models could be misused by hostile states or other actors if released without prior vetting. It is framed within a broader policy effort: an executive order signed earlier in June 2026 that directed agencies to set up a voluntary testing-and-evaluation process for powerful new AI models before public release, The Verge reported.

The current, targeted approach — approving customers one by one rather than requiring a formal license — is described as a compromise after an earlier, stricter draft order met industry pushback.

OpenAI's stated position

Altman did not present the arrangement as permanent or preferred. In the memo quoted by reporters, he said OpenAI had made clear to the government that "this is not our preferred long term model" and that the company would work toward "a more sustainable approach for future releases," while describing the staggered rollout as the fastest route to a broad release. newsparlor could not independently confirm the memo's wording.

Why it matters

If accurate, the episode would mark an unusual moment: a US administration asking an American AI company to hold back a commercial product before it reaches the public, under a process described as voluntary but carrying clear government leverage. Supporters of such review argue that frontier AI systems warrant security checks given their potential for misuse, often citing competition with China; critics worry about government influence over which products reach the market and on what terms. With neither OpenAI nor the White House commenting on the record, key details — including how long the preview will last — remain unconfirmed.