White House limits OpenAI's GPT-5.6 launch
Reporting on June 25 says the Trump administration asked OpenAI to release GPT-5.6 to government-approved customers first, approving access one by one.
The Trump administration has asked OpenAI to hold back the public release of GPT-5.6, its next frontier model, according to reporting on June 25. Instead of a broad launch, OpenAI plans to share the model with a small set of government-approved partners first. CEO Sam Altman told staff the administration planned to approve “access customer by customer” while the model sat in preview. OpenAI hopes a wider release follows a couple of weeks later.
Two White House bodies made the request: the National Cyber Director’s office and the Office of Science and Technology Policy. It follows an executive order Trump signed in early June directing some AI companies to submit new models for government testing before release. Officials reportedly view GPT-5.6 as roughly on par with Anthropic’s Mythos, the model the same administration effectively pulled from the market through an export-control order earlier this month.
What it means for you
This is the second frontier model in a month to ship under government gatekeeping rather than a normal launch. For operators, the lesson is access risk: the strongest model may arrive late, gated, or not at all in your region. That is the same exposure behind the export controls that took top Claude models offline, and a good reason to read what to do when your AI model gets switched off before you build a workflow around a single vendor.