GPT-5.6 Delayed: Trump Administration Asks OpenAI to Slow Its Next Release

The Trump administration has asked OpenAI to stagger the release of GPT-5.6, citing security concerns. According to a report by The Information, OpenAI CEO Sam Altman told employees on Wednesday that GPT-5.6 will launch in a limited preview restricted to a small group of enterprise customers. The federal government will reportedly review and approve access for each customer individually during that preview window. The arrangement appears more favorable to OpenAI than a similar deal the Trump administration struck with competitor Anthropic.

What happened

OpenAI CEO Sam Altman revealed during an internal company Q&A on Wednesday that the next major model, GPT-5.6, will not receive a standard public launch. Instead, according to reporting by The Information, Altman told employees the model will first roll out in a limited preview available only to a select group of enterprise customers.

The reason: the Trump administration raised security concerns and asked OpenAI to slow the release down. During the preview period, the federal government itself will reportedly decide which customers get access, reviewing requests on a case-by-case basis.

The source of the request is notable. This is not OpenAI choosing to run a cautious beta on its own terms. The pace and scope of the rollout are being set, at least in part, by the White House.

Why it matters

For businesses planning around GPT-5.6 capabilities, this adds real uncertainty. A limited enterprise preview controlled by government approval means most companies will not simply be able to sign up and get access at launch.

A few things worth paying attention to:

• Access is not guaranteed at launch. Even enterprise customers may need to wait for federal sign-off.
• The approval process is opaque. There is no public criteria for how or how quickly the Trump administration will approve customers.
• Anthropic got a harder deal. According to the same reporting, the arrangement offered to OpenAI is more favorable than what the Trump administration gave Anthropic, suggesting these negotiations are happening model by model and company by company.
• This sets a precedent. If this pattern holds, future frontier model releases could routinely involve government-gated access periods before broader availability.

For most small and mid-sized businesses, the practical impact depends on how long the preview lasts and how quickly access opens up beyond that first group. But the direction of travel here is worth noting: government bodies are now actively shaping when and who gets access to leading AI models.

Our take

From where we sit, the most interesting detail is not the delay itself. It is the mechanism. A government body approving AI model access on a customer-by-customer basis is a meaningful structural shift, not just a one-off request.

OpenAI gets to frame this as cooperation with national security interests, which costs them relatively little if the preview period is short and enterprise demand holds. But a precedent where federal approval gates who can use a new model has long-term implications for how AI capabilities flow to businesses, researchers, and developers.

It also raises a practical question for anyone building on OpenAI’s API: if GPT-5.6 access is rationed at launch, how long does it take before your tier gets included? That timeline is currently unknown, and OpenAI has not published criteria for how the preview group was selected.

The Anthropic comparison is worth watching too. If rival companies face stricter or looser access conditions than OpenAI, that shapes competitive dynamics in ways that have nothing to do with the technology itself.

What to do about it

If your business is planning a project that depends on GPT-5.6 capabilities, do not build that timeline around a standard launch date. Track The Information and OpenAI’s official announcements for any update on when the preview expands. For now, design around what is currently available and treat GPT-5.6 access as a variable, not a given.

Source: The Verge · AI