TECH NEWS – OpenAI has begun rolling out GPT-5.6, but the new model family will not initially reach the general public. At the request of the Trump administration, access is temporarily limited to a small group of vetted partners while Washington develops a process for examining the cybersecurity risks of advanced AI models. OpenAI is cooperating, but has made clear that it does not want this kind of government-controlled access process to become the long-term default.
OpenAI began a limited preview of its GPT-5.6 model family on Friday. The release includes three variants: Sol, the most capable version; Terra, a balanced model designed for high-volume work; and Luna, a faster and more affordable option. The company says the models are particularly strong at coding, cybersecurity tasks, biology-related work, and long-running agentic workflows that require sustained focus across multiple steps.
The rollout is not happening in the usual way, however. At the request of the US government, GPT-5.6 is currently available only to a small group of trusted partners, with access being approved on a case-by-case basis during the preview period. Axios reports that around twenty companies are taking part in the early programme, while Reuters says federal officials are approving customer access individually.
The main concern is the model’s stronger cybersecurity capability. Earlier this month, the US government established a framework allowing it to examine the national-security risks posed by the most advanced AI systems before they receive broad public release. This makes the current situation especially notable, because it is one of the first cases in which Washington has requested a limited, phased launch for a specific American AI model before it becomes widely available.
OpenAI has stressed that GPT-5.6 Sol is more useful for identifying and fixing vulnerabilities than for carrying out reliable end-to-end cyberattacks, and that it does not cross the company’s own critical cyber-risk threshold. Even so, the company acknowledges that unforeseen risks may emerge when the model is combined with other tools, which is why it is pairing the release with stronger safeguards and a phased deployment.
“We don’t believe this kind of government access process should become the long-term default,” OpenAI said. The company describes the restriction as a temporary step that could lead to broader availability in the coming weeks while it works with the administration on a repeatable process for future model launches. This does not mean the US government has permanently taken control of OpenAI’s release decisions, but GPT-5.6 clearly shows that access to the most advanced AI systems is now becoming a direct matter of government policy as well as business and safety.
Leave a Reply