OpenAI’s next major model launch is moving more slowly than expected, after the Trump administration reportedly pushed the company to limit early access to GPT 5.6. Instead of a broad public rollout, the model is beginning with a small group of trusted partners while government review continues.
The unusual arrangement highlights how closely advanced AI releases are now tied to national security concerns. For OpenAI, the immediate issue is not only when GPT 5.6 reaches users, but also how much of the launch process must now pass through government scrutiny.
Government review before wider release
According to The Information, OpenAI CEO Sam Altman told staff that the government would “approve customer access one by one” for GPT 5.6. If the limited phase goes well, the company expects the model to become more widely available within a few weeks.
The model is being reviewed because of security concerns, and the Office of the National Cyber Director as well as the Office of Science and Technology Policy have reportedly requested access. That means GPT 5.6 is not going straight to the public, even though OpenAI has already presented its plans and model capabilities to the US government.
OpenAI said it is following that framework by starting with a restricted preview and then expanding after testing and coordination with partners. The company also said it does not see this as an ideal long-term standard, because it could keep its strongest tools away from users, developers, companies, cybersecurity defenders, and global partners who need them.
Why GPT 5.6 drew attention
GPT 5.6 arrives as a three-model family: Sol, Terra, and Luna. OpenAI says the lineup brings notable gains in coding, biology, and cybersecurity, which is also why safety has become such a major issue around the release.
Sol is the main model and will use what OpenAI calls its “most robust safety stack to date,” with stronger protections against high-risk activity, sensitive cyber requests, and repeated misuse. Terra is the balanced option, while Luna is positioned as the faster and more affordable model.
| Model | Positioning | Input Price | Output Price |
|---|---|---|---|
| Sol | Main model | $5 per 1M tokens | $30 per 1M tokens |
| Terra | Balanced option | $2.50 per 1M tokens | $15 per 1M tokens |
| Luna | Fastest and cheapest option | $1 per 1M tokens | $6 per 1M tokens |
The pricing table above shows how OpenAI is splitting the family into premium, balanced, and low-cost choices. For now, however, those prices matter less than the fact that access is still confined to a limited preview.
A wider rollout is still the goal
OpenAI says the limited launch is meant to be the fastest path toward broader availability in the coming weeks. At the same time, the company said it is working with the Trump administration on an executive-order cybersecurity framework and a repeatable process for future model launches.
The situation makes GPT 5.6 one of the clearest examples yet of how powerful AI systems can face direct government involvement before public release. It is also a sign that future launches may face more scrutiny if the models are viewed as sensitive enough to warrant early review.
For now, GPT 5.6 remains in controlled access, with broader availability dependent on the outcome of the restricted preview and the approval process that is unfolding alongside it.
