OpenAI is limiting preview access to its latest GPT models at the request of the US government.<\/p>\n
The company said it previewed the capabilities of its new series of GPT-5.6 models, which includes Sol, Terra, and Luna, to the US government ahead of Friday’s launch.<\/p>\n
“At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly,” the company said.<\/p>\n
OpenAI said they were complying as part of its earlier agreement with the Defense Department, which allows the Pentagon to use its AI models.<\/p>\n
Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.https:\/\/t.co\/OoM83SyISN<\/a><\/p>\n \u2014 OpenAI (@OpenAI) June 26, 2026<\/a> <\/p>\n OpenAI said the new models will become available to the wider public in the coming weeks.<\/p>\n The limited GPT-5.6 release comes after OpenAI’s chief rival,\u00a0Anthropic, was forced to revoke access<\/a> to its latest models, Mythos and Fable, on June 12 following the Trump administration’s imposition of broad export controls. Anthropic has yet to restore access to either model.<\/p>\n The debate around AI model safety has surged since Anthropic announced the limited release of Mythos Preview in April. At the time, Anthropic said it would release its most powerful model only to a limited number of cybersecurity firms and trusted companies due to Mythos’s advanced hacking<\/a> capabilities. The disclosure that an AI model could easily find vulnerabilities in existing software and cyber infrastructure spooked investors and world leaders.<\/p>\n When Anthropic initially released Fable 5<\/a> on June 5, it said the model was in the “Mythos-class” but had enough safeguards in place to make it suitable for a broader public release.<\/p>\n