OpenAI is about to release its next model, GPT-5.6, and for the first time the rollout is being shaped by the federal government. According to The Information, the Trump administration asked OpenAI to stagger the launch, releasing the model to a small set of approved customers first while the government signs off on access one customer at a time. CEO Sam Altman relayed the request to staff this week.
What OpenAI agreed to
Instead of a broad public launch, GPT-5.6 will go out as a limited preview. During that window, the government will be "approving access customer by customer," according to Altman's note to employees. The request came out of conversations with two federal offices: the Office of the National Cyber Director and the Office of Science and Technology Policy. Neither OpenAI nor the White House has publicly confirmed the details or said how long the controlled phase will last, though some accounts suggest a wider release could follow in a couple of weeks.
Why the government stepped in
The stated reason is cybersecurity. The concern is that a model capable enough to find and exploit software weaknesses could do real damage if it is handed to the wrong users on day one. Rather than block the model, the government is asking for a slow, supervised rollout so it can watch how the most capable version behaves before it reaches everyone.
A bigger pattern
This did not come out of nowhere. The government ordered Anthropic to pull its Fable 5 model earlier this month, and OpenAI has since said it would comply with an executive order encouraging companies to let the government review frontier models before release. Taken together, the moves point to a new norm: the most capable AI systems are starting to be treated like sensitive technology, with a government check between "the model is finished" and "anyone can use it."
Supporters framed it as responsible risk management. Critics, including the tech crowd that wants AI to move as fast as possible, called it government overreach that could slow the entire industry. Sentiment online ran mostly negative, but the underlying facts are not really in dispute: the rollout is being managed, and the government has a hand on the dial.
I mostly wanted to put this on your radar: GPT-5.6, OpenAI's next model, probably isn't reaching the rest of us until July, maybe later.
It also points to how model releases work now. New models aren't going to just drop for everyone the day they're ready. They'll be phased in slowly, a few approved customers at a time, while the security reviews play out. Gone are the days of a big model launching to all of us at once, which is kind of a bummer.
There's a fair case for caution, since the newest models are good enough at finding software holes that a slow rollout isn't crazy. There's also the worry that "we need to vet it first" becomes a quiet say over who gets the best tools, and when.
Either way, if you've been waiting on the next big upgrade, settle in. The stretch between a model being finished and you actually getting to use it is now its own waiting game.
On a lighter note:
BREAKING: Google flagged its next Gemini model for mandatory safety review as well. The government assured them that won't be necessary.
— Per Simmons (@per_simmons_) June 25, 2026
This is a joke, of course. Sorry for the low blow, Google.