WASHINGTON, July 8 (DC Times Online) — OpenAI is not opening its newest GPT-5.6 model to everyone yet. Instead, the company said it is giving access to a small group of trusted partners first while the Trump administration reviews possible cybersecurity risks.
For most users, that means the model is still in preview, not in broad public release. OpenAI said wider access is expected in the coming weeks if the limited rollout goes as planned.
What OpenAI is releasing
OpenAI said the newest model in its GPT-5.6 line is being shared only with a limited group of partners for now. The company described the model as its strongest yet and said it is better at helping people find and fix software vulnerabilities than at carrying out cyberattacks.
In its public statement, OpenAI also said the model did not cross its internal “Cyber Critical” threshold, which is the company’s own line for especially risky cyber capability.
The company said the preview is temporary and that it does not want government access to become the long-term default.
Why the White House is involved
The rollout sits inside a broader federal effort in Washington to look at advanced AI models before they are widely released.
That effort traces back to a June executive order signed by President Donald Trump. The order set up a voluntary federal review framework for powerful AI systems and described a review period that can last up to 30 days. The White House has said the goal is to help assess possible threats to financial systems, national security and other critical systems.
OpenAI said the limited preview came after a request from the White House. Reporting has said the offices involved include the Office of the National Cyber Director and the Office of Science and Technology Policy.
The White House has also said it continues to work with frontier AI labs on shared ways to handle the risks of scaling the technology.
What “limited preview” means
A limited preview is not the same as a general launch. It means only a small, approved group gets early access while the company tests the model and watches for problems.
OpenAI said those early users are trusted partners, and that their participation has been shared with the government. The company has not publicly named the partners.
That kind of staged release is meant to give the company and the government time to study whether the model can be misused. In this case, the concern is cyber risk: whether a powerful model can help defenders spot flaws, but also make it easier for attackers to exploit them.
Why cyber risk is the focus
OpenAI framed the model as useful for defense work. But powerful models can also lower the barrier for harmful tasks, such as writing malicious code, finding weaknesses in software or automating parts of an attack.
That is why the administration’s review process is centered on cyber issues. The White House framework is meant to let officials and companies examine whether a model could pose dangers before it reaches a wider audience.
This is also why the release matters beyond OpenAI. If the government pushes for more previews and more review time, that could become a model for how future AI systems are handled in Washington.
What users should expect next
For now, the main change is simple: the newest GPT-5.6 model is not broadly available yet.
OpenAI said the preview period is short-term and that broader access should follow in the coming weeks. If that happens, the model could reach more users soon. If the review or testing uncovers problems, the launch could stay limited longer.
The bigger story is not just one product. It is the new role Washington is trying to play in checking powerful AI systems before they spread more widely.
That puts the debate about AI safety, cybersecurity and release speed squarely in the federal capital, where the next step for one company could influence how other AI models are reviewed too.
