OpenAI says access to its new GPT-5.6 model is limited at the request of the US government

In a significant move, OpenAI has announced that access to its innovative GPT-5.6 models will be restricted at the behest of the US government. Ahead of the official launch scheduled for Friday, the company provided a preview of its latest models—Sol, Terra, and Luna—to officials in Washington. OpenAI stated that it is initiating a limited preview for a select group of trusted partners, a decision made in cooperation with the government. This approach is part of a prior agreement with the Department of Defense, which permits the military to utilize OpenAI's advanced models. The new GPT-5.6 Sol model is designed as a cutting-edge frontier tool, while Terra serves as a balanced option for everyday tasks, and Luna is tailored for high-volume work with speed and efficiency. OpenAI emphasized that this limited access should not become a long-term standard, arguing that it restricts essential tools from users, developers, and enterprises who could benefit from them. The company expressed hope that this temporary measure would facilitate broader availability in the near future, as it collaborates with the Administration on developing a framework for cybersecurity executive orders and a more consistent process for future model releases. OpenAI has indicated that the general public can expect access to the new models in the upcoming weeks. In a related context, Anthropic recently withdrew access to its latest models, Mythos and Fable, after security concerns were raised, prompting the company to limit access for foreign users as well.