GPT-5.6 Rollout Delayed: US Government Reportedly Seeks to Approve Customers One by One Before Wider Release

OpenAI has reportedly delayed the wider release of GPT-5.6 after the US government requested a staggered rollout that would require individual customer access approvals during an initial preview phase.

The move, which was reportedly agreed by OpenAI, means the latest model will be made available to a limited group of partners before any global release.

The move may have been implemented after growing scrutiny around the deployment of advanced artificial intelligence systems, with governments and technology companies increasingly focused on how powerful models are introduced to users.

According to reports, the Trump administration's concerns about security risks led officials to ask OpenAI to slow the release of GPT-5.6 rather than make the system widely available immediately.

GPT-5.6 Rollout Faces New Restrictions During Preview Phase

To recall, OpenAI had been preparing to introduce GPT-5.6 as its latest frontier model, but the reported discussions with government officials changed the path towards a more controlled launch. OpenAI chief executive Sam Altman reportedly told employees during a Wednesday question-and-answer session that the company would release the model first to a small group of partners as part of a limited preview.

The reported decision followed a similar approach to that of rival artificial intelligence company Anthropic, which shared its Mythos model with select partners rather than immediately opening public access. The comparison has highlighted a possible shift in how major AI laboratories handle the release of systems with advanced capabilities, particularly those involving cybersecurity or other sensitive applications.

In a Thursday memo to staff, Altman reportedly said the government would be 'approving access customer by customer during this preview period' for GPT-5.6. The arrangement means organisations seeking access would need individual approval before they could use the model during the early stage of deployment.

The details of the approval process, including which customers would qualify and how decisions would be made, have not been publicly outlined.

A More Cautious AI Release Model

The reported delay comes as governments and AI companies are discussing how quickly powerful new AI systems should be released. While companies want to make their latest models available to users, they also face concerns that advanced AI could be misused if access is not carefully controlled.

Reports suggest the US government's involvement in the GPT-5.6 rollout could influence how other AI companies release future models. Instead of companies deciding alone when and how their most powerful systems become available, governments could play a bigger role in reviewing who gets access.

OpenAI has not publicly explained the exact security concerns behind the reported restrictions, nor have the government's specific requirements been revealed.

But the reported move follows similar steps taken by other AI companies. Anthropic previously granted limited access to its Mythos model rather than releasing it publicly straight away, suggesting that some developers are already adopting a more cautious approach when launching advanced AI systems.

What To Expect From GPT-5.6 So Far

GPT-5.6 is expected to be less about a flashy new chatbot feature and more about making AI feel like a much more capable assistant that can handle bigger, more complicated tasks.

While OpenAI has not officially revealed a full list of improvements, reports and industry expectations suggest the next major upgrade could focus on better reasoning, stronger coding abilities, improved reliability, and more advanced 'agent' capabilities, in which the AI can complete tasks with less step-by-step guidance from users.

One of the biggest changes people are watching for is whether GPT-5.6 can move beyond simply answering questions and start acting more like a digital worker. That could mean helping users complete longer projects, researching information, analysing documents, writing and testing code, or managing complex workflows. Instead of asking an AI dozens of small questions, users could give it a larger goal and let it work through the steps needed to reach a result.

Another expected improvement is a larger context window, which refers to how much information the AI can understand and remember during a conversation or task. Some unconfirmed reports have suggested that GPT-5.6 could handle much more information than previous models. If accurate, that could allow professionals to upload huge documents, entire software projects, or large collections of research material and have the AI analyse them together.

However, OpenAI has not confirmed these figures.

For people and businesses waiting to use GPT-5.6, the timing of its release remains uncertain. Reports suggest access during the preview stage will depend on individual approval before OpenAI allows global availability.