OpenAI plans to release GPT-5.6 to a small group of partners before any public launch, and the federal government will sign off on who gets in, one customer at a time. CEO Sam Altman told staff about the staggered plan during an internal Q&A on Wednesday, June 25, according to The Information, which broke the story.
Washington picks who goes first
The request came from two White House offices: the Office of the National Cyber Director and the Office of Science and Technology Policy. In a memo to employees, Altman wrote that the government would be "approving access customer by customer during this preview period," with a broader rollout he hoped would follow "a couple of weeks later."
Read that twice. A federal review board deciding, account by account, who can use a commercial chatbot is not how OpenAI has shipped anything before. Altman framed it as the fast path, not the forced one, telling staff the staggered approach was the quickest route to a broad release. Maybe. It is also the kind of thing you say when the alternative is worse.
Altman was blunt that this won't be the permanent arrangement. Axios reported that OpenAI had been working with the administration on the release since before the Anthropic episode, so the cooperation predates the headlines.
"We've made clear to the U.S. government that this is not our preferred long term model." That's from Altman's memo, and the phrase "long term model" is doing a lot of quiet work. He's fine with it for now.
The Lutnick call
Commerce Secretary Howard Lutnick talked to Altman the same Wednesday. Per Axios, Lutnick wanted assurance that every relevant part of the government had tested and approved the model before launch. A source described the intervention as routine for a model of this caliber rather than a sudden crackdown, saying GPT-5.6 has "Mythos-like" capability, particularly around cybersecurity.
Take the "routine" framing with some salt. It's coming from a source defending the arrangement, and no one has said what the security criteria actually are or how a customer qualifies. International clients have no idea whether US sign-off applies to them either.
So much for voluntary
Here's the awkward part. Trump signed an executive order earlier this month directing agencies to build a voluntary testing protocol for frontier models before release. Voluntary. A customer-by-customer government gate looks a lot like a licensing regime wearing a voluntary costume, and the gap between the order's language and what's happening to GPT-5.6 is hard to miss.
The precedent is Anthropic. The company shared its Mythos model with select partners in April rather than launching it widely. Then on June 12, a Commerce Department export control directive forced Anthropic to pull Fable 5 and Mythos 5 entirely offline to block foreign access. That didn't exactly work out as a clean template, which makes OpenAI's willingness to walk a similar road worth watching.
OpenAI and the White House haven't publicly confirmed the reports. The broader GPT-5.6 release is expected roughly two weeks after the preview period, assuming the approval process goes smoothly. Earlier timing for the model had already slipped toward July.
