The Trump administration blocked a global rollout, forcing the AI firm to clear every user one by one with federal officials.
OpenAI restricted GPT-5.6 to US government-approved partners on June 26, 2026, after the Trump administration cited national security concerns over the model’s advanced capabilities.
On Friday, OpenAI restricted its new GPT-5.6 model to a narrow circle of US government-approved partners before any wider release, after the Trump administration raised national security concerns about the model’s advanced capabilities.
It was the first time Washington had moved preemptively to gate an American AI company’s model before it reached the public. But for most users worldwide, Friday’s announcement changed nothing at all.
OpenAI’s GPT-5.6 family arrives in three versions, Sol, Terra, and Luna, engineered to handle distinct tiers of intelligence, everyday performance, and processing speed. The company confirmed it is running “a limited preview” of the series, with broader access pledged in coming weeks.
Two White House offices drove the request: the Office of the National Cyber Director and the Office of Science and Technology Policy. Neither offered a public timeline for how long the government’s approval process would run.
Wednesday afternoon, Sam Altman sat down with Commerce Secretary Howard Lutnick, who wanted confirmation that the relevant parts of the federal government had reviewed and cleared the model before any rollout began.
Altman disclosed the restriction to employees during an internal question-and-answer session on Wednesday, June 25. He later confirmed in a memo that Washington would be “approving access customer by customer during this preview period.” And he told staff he hopes to release GPT-5.6 more broadly just “a couple of weeks later.”
Access is rolling out initially to twenty partners. Amazon’s Bedrock software platform is among the entry points. No specific public launch date appeared in Friday’s announcement.
In the memo, Altman told employees plainly: “We’ve made clear to the U.S. government that this is not our preferred long-term model and will work with them and others in industry to achieve a more sustainable approach for future releases.”
The GPT-5.6 episode follows a far messier confrontation between the White House and Anthropic. On June 12, the Trump administration issued an export control directive ordering Anthropic to pull its Fable 5 and Mythos 5 models entirely offline after roughly ninety minutes’ notice.
Because restricting access by nationality proved technically unworkable at that speed, Anthropic shut both models down for every user everywhere. Yet the episode clarified exactly why Washington moved on OpenAI when it did.
Both OpenAI and administration officials regard GPT-5.6 as matching Anthropic’s Mythos in raw capability, particularly in cybersecurity applications. That equivalence, confirmed to Westtrybe, was central to the government’s calculus.
Earlier this month, President Trump signed an executive order titled “Promoting Advanced AI Innovation and Security,” directing multiple agencies to establish a voluntary pre-release testing protocol for AI companies. No binding federal framework currently governs this process.
A White House official told reporters the administration continues “to collaborate with frontier AI labs to develop shared approaches for addressing the challenges of scaling this technology.” OpenAI did not specify which international markets would gain access once the domestic preview concludes.
For now, GPT-5.6 remains a product the world can read about but cannot open. Its rollout is being shaped less by engineering timelines than by the national security arithmetic being worked out somewhere between San Francisco and Washington.
