WASHINGTON, July 8 (DC Times Online) — OpenAI has put a temporary gate in front of its newest AI model.
The company says the model, which it identified as GPT-5.6 Sol, is being released first to a small group of trusted partners instead of the public. OpenAI said the limited rollout is happening while the Trump administration reviews the cybersecurity risks of powerful AI systems.
For most users and businesses, the main takeaway is simple: the model is not broadly available yet. OpenAI says wider access is expected in the coming weeks if the preview period goes well.
What did OpenAI change?
OpenAI said it is starting with a limited preview rather than a full release. The company said the goal is to share the model first with partners it trusts, and that those partners have been shared with the government.
OpenAI also said the model is its strongest yet and that it is better at helping people find and fix software vulnerabilities than at carrying out cyberattacks. The company said the model does not cross its internal “Cyber Critical” threshold, a label it uses for especially risky cyber capabilities.
Why is the White House involved?
The Trump administration has created a federal review framework for advanced AI models. In a June executive order, President Donald Trump set up a voluntary process for government vetting before release, with a review window of up to 30 days.
The White House has said the process is meant to help federal officials assess possible risks to national security, financial systems and other critical systems. The offices named in reporting as part of that process are the Office of the National Cyber Director and the Office of Science and Technology Policy.
OpenAI said it made the limited release at the request of the White House. At the same time, the executive order is described as voluntary, and the reporting does not make clear whether this specific pause was required or simply requested.
What does a limited preview mean in practice?
A limited preview means the model is not yet available to the general public or to all ChatGPT users. Instead, OpenAI is letting a small group of approved partners test it first.
That matters because advanced AI models can be used for both defense and offense. In cybersecurity, a model that can spot weak points in code may help companies patch systems faster. But the same kind of capability can also be misused to look for flaws to attack.
OpenAI says its current release plan is meant to balance those risks with access for developers and defenders who may need the tool. The company said it does not want a government-access process to become the long-term default.
Why does this matter beyond one product?
This rollout shows how AI releases are becoming a policy issue, not just a product launch. When a model is powerful enough to raise cyber concerns, companies may face pressure to slow down, limit access, or send it through government review.
For Washington, the broader question is how far federal oversight should go before frontier AI systems reach the public. For users, the practical effect is less abstract: the newest tool may arrive later, in smaller batches, and with more checks attached.
OpenAI says the wider rollout should follow after the preview period, which it describes as short-term. Until then, the model remains in controlled testing rather than general release.
