The US government has stepped in to directly control the release of OpenAI’s next major model, GPT-5.6, marking a significant shift in how frontier artificial intelligence reaches the public.
The Trump administration has asked OpenAI to limit the release of GPT-5.6 to select government-approved partners before any wider launch, according to The Information. The move comes amid security concerns over the model’s advanced capabilities, which reportedly match the threshold set by Anthropic’s Mythos 5.
Under the staggered rollout plan, OpenAI will preview GPT-5.6 to a small group of vetted partners. Crucially, the government will approve access on a “customer by customer” basis, giving Washington direct oversight over who gets to use the model first.
In an internal memo to employees, OpenAI CEO Sam Altman described this arrangement as the best path to getting GPT-5.6 released at all. He indicated that a general public release would likely follow “a couple of weeks later.”
Altman also told staff that OpenAI has made clear to the White House that this is not the preferred long-term approach. The company plans to work toward a more sustainable release model for future iterations beyond GPT-5.6.
A growing pattern of government intervention
This is not an isolated incident. First came government scrutiny around Anthropic’s Claude Fable 5, followed by the Mythos 5 restrictions. Now GPT-5.6 is subject to the same treatment. The pattern is becoming unmistakable.
The US government is actively deciding when and how frontier AI systems reach the public. The framing is national security, not restriction, but the precedent is now firmly established. As models continue to approach Mythos-level capability, government sign-off may become a standard step before any major AI release.
What this means for the industry
For AI companies, this introduces a new and unpredictable variable into product timelines. A model might be technically ready for release but face weeks or months of government review before it can reach users. That creates uncertainty around development roadmaps, commercial expectations, and competitive positioning.
For the broader technology sector, the question becomes whether this pattern extends beyond frontier AI models. If the government is willing to gatekeep advanced AI capabilities, could other transformative technologies face similar controls?
OpenAI’s GPT-5.6 release is now effectively on a government-mandated timeline. The company’s cooperation suggests a pragmatic recognition that fighting the restriction would only delay things further. Whether this collaborative approach sets a healthy precedent, or simply normalises government oversight of AI releases, remains to be seen.
One thing is clear: the era of unchecked frontier AI launches is over. From this point forward, the most capable models will clear both technical and regulatory hurdles before reaching the public.