OpenAI is about to release GPT-5.6, its next AI model, but you will not be able to try it right away. The U.S. government will decide, one customer at a time, who is allowed to use it before any wider launch.
The Gist
- GPT-5.6 starts with a small group of users picked by the White House.
- A broader release is planned a couple of weeks later if the first phase goes smoothly.
- The setup follows another model, Fable 5, that was switched off by the government.
Why You Cannot Just Try GPT-5.6 Yet
If you are used to logging in and trying the latest ChatGPT model the day it drops, GPT-5.6 will break that habit. OpenAI is launching it under a slow rollout requested by the U.S. government, with each customer reviewed before being added to the list.
Sam Altman, OpenAI’s CEO, told employees about the plan during an internal Q&A on Wednesday. He admitted the company is going along with the process for this release, but said it is not the kind of long-term setup OpenAI wants for future models.
In simple terms, GPT-5.6 is a frontier model. That means it sits at the very edge of what AI can do today, which also means it can do things that worry safety experts. Writing malware, for example, or running attacks online without a human at the wheel.
The U.S. government is treating this kind of model less like a software update and more like a sensitive tool. The White House wants to know who is using it before that user gets in, not after something goes wrong.
For everyday users, this matters in a very practical way. The version of ChatGPT inside your phone or your browser is not going to jump to GPT-5.6 the moment it launches. The early access is reserved for a few partners, and most of us will see the upgrade later.
How This New Process Works
The screening goes through two federal offices. The Office of the National Cyber Director looks at the cybersecurity risks, and the Office of Science and Technology Policy looks at what the model can newly do. Both work directly with OpenAI to pick the first authorized customers.
Howard Lutnick, the U.S. Commerce Secretary, reportedly called Sam Altman to insist on adding more agencies to the review. What started as a single White House check is now becoming a multi-desk procedure, which is a sign that Washington is taking this seriously.
The whole thing is rooted in an executive order signed by Donald Trump in early June 2026. That order asks select AI companies to voluntarily send their new models to the government for testing before launching them to the public.
“Voluntarily” is a soft word in this context. Anthropic, OpenAI’s rival, was forced to take down a model called Fable 5 after the government said it was a national security threat. Anthropic had even released its earlier Mythos model carefully, through a program called Project Glasswing, but that still was not enough.
OpenAI has clearly watched this happen. By accepting a customer by customer review up front, the company avoids the worst case: launching widely, then being told to pull the plug a few weeks later.
Keep learning on AI Noobies:
- Claude Tag: Anthropic’s Always-On Slack Assistant
- AI Loops: How Bots Code While You Sleep
- iOS 27 AI Features: 8 Things Your iPhone Will Do
What This Means For Everyday AI Users
In the short term, you should expect a more boring launch than usual. No flashy public day-one, no immediate upgrade in your ChatGPT app, just a quiet rollout to a few approved companies that will likely test GPT-5.6 in highly controlled settings.
Once that preview phase ends without issues, OpenAI plans to widen access a couple of weeks later. Even then, the broader release may come with extra guardrails, especially around what businesses can build on top of the model.
In the medium term, this is the early shape of how frontier AI may be governed in the United States. A model like GPT-5.6 is no longer just a product. It is a tool the government wants a seat at the table for, and OpenAI has agreed, even if reluctantly.
If you are a beginner watching the AI space, the takeaway is simple. The biggest models are starting to look less like apps and more like regulated infrastructure. Who gets to use the most powerful AI, and when, is becoming a decision that involves the U.S. government, not just the companies that build the models.
For now, your ChatGPT will keep working as usual. The GPT-5.6 you read about today is being parceled out behind closed doors, and the wider world will only meet it when Washington signs off.
Follow the story on AI Noobies.
About the Author
