On June 2, 2026, President Trump signed an executive order on AI oversight. It set up a framework for the federal government to vet the most advanced AI systems for national security risks. The vetting could take up to 30 days before a public release. Participation was described as "voluntary".
Twenty-four days later, the "voluntary" framework became something else entirely. On June 25, Axios reported that the White House had asked OpenAI to phase the release of its next-generation model, GPT-5.6. It was the first time the U.S. government had ever preemptively asked an American AI company to restrict the launch of a model before release.
The "voluntary" framework had become, in practice, something far more powerful.
The Request
The White House's Office of the National Cyber Director and Office of Science and Technology Policy asked OpenAI to limit the rollout of GPT-5.6. The administration was still building a framework for testing and evaluating the security of new models. It believed the model's capabilities were comparable to Anthropic's Mythos model. That model had already triggered an export control order. The government intervened because GPT-5.6 had "Mythos-like" capability.
OpenAI CEO Sam Altman agreed. In a memo to employees, he said GPT-5.6 would debut as a preview. It would be available only to a small number of partners. The U.S. government would approve access "customer by customer".
Altman also made his position clear: "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".
Altman had discussed GPT-5.6 with Commerce Secretary Howard Lutnick on June 24. Lutnick wanted to be sure all relevant parts of the government had tested the model. Only then could it see a broad release.
The Release
On June 26, OpenAI announced that GPT-5.6 Sol — its most powerful model yet — would be available only to a "small group of trusted partners" approved by the Trump administration. The full GPT-5.6 lineup included three models: Sol (flagship), Terra (balanced, everyday use), and Luna (faster, lower-cost). All three were restricted.
The preview was limited to partners "whose participation has been shared with the government". OpenAI called it a "short-term step" on the "path to broader availability in the coming weeks".
The company also made its position public: "We don't believe this kind of government access process should become the long-term default". "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them".
The Precedent
The request was not made in a vacuum. It followed the Anthropic precedent.
On June 12, the Commerce Department issued an export control order against Anthropic. It demanded that all foreign nationals be blocked from accessing Fable 5 and Mythos 5. Anthropic, unable to filter users by nationality, took both models offline entirely.
The message to the industry was unmistakable. If your model is powerful enough, the government can effectively ban it — even after release.
David Sacks, co-chair of Trump's Council of Advisors on Science and Technology, had warned earlier that the world had "no choice but to take the cyber threat associated with Mythos seriously". But he also cautioned that without real-world demonstrations of Mythos threats, the company risked a "serious credibility problem".
Stanford cybersecurity expert Alex Stamos pushed back. "Pretty much nobody in the cybersecurity industry believes that there's any factual basis for this action," he said. He noted that he reviewed Amazon's analysis of Fable and "didn't find any risks that aren't present with other publicly available AI models, including those made in China".
The Irony
The irony is difficult to miss.
On June 2, the administration promised a "voluntary" framework for AI oversight. By June 26, it was approving customers one by one.
On June 12, the Commerce Department effectively banned Anthropic's models. By June 26, the White House was restricting OpenAI's models before they were even released. The "voluntary framework" had become a de facto licensing regime.
Dean Ball, a former White House AI adviser, argued that the executive order had created a "de facto involuntary licensing regime" for frontier AI. He warned that without clearly defined safety standards, the government could cause endless launch delays. That could give China an advantage in the AI race and jeopardize billions in AI infrastructure investment.
The Full Release
On July 9, OpenAI announced that GPT-5.6 Sol, Terra, and Luna would launch publicly. The decision followed additional testing and meetings between the company and government officials.
OpenAI stated it had "consulted extensively with the U.S. government throughout the GPT-5.6 release process". The administration had reportedly requested a phased release instead of a full public launch. It cited the need for additional verification of national security risks.
The approval came with a clarification. A White House official disputed that the administration gave OpenAI a "green light" because such permission "isn't necessary." "No such permission is required or granted," the official said. Decisions on the timing and scope of releases "rest entirely with the companies". The official pointed to Trump's June 2 executive order. It bars any mandatory federal licensing or preclearance for releasing AI models. The administration was maintaining the fiction of a "voluntary framework". Even so, it had already shown the power to restrict — and effectively approve — frontier AI releases.
The New Dynamic
The industry has crossed a threshold. The U.S. government has now shown that it can — and will — intervene in frontier AI releases. It acts both before and after launch.
The executive order was described as "voluntary," but the actions speak louder than the language. Anthropic's models were shut down. OpenAI's models were restricted to government-approved customers. The "voluntary" framework is, in practice, a licensing regime — administered without clear standards, transparent processes, or statutory authority.
OpenAI has made its position clear: "We don't believe this kind of government access process should become the long-term default". But the precedent is set. The question is no longer whether the government will intervene in AI releases. It is how far the intervention will go. The other question is whether the industry can build a more sustainable framework before the government builds it for them.
Sources
- White House Fact Sheet on Executive Order "Promoting Advanced Artificial Intelligence Innovation and Security" (June 2, 2026)
- Axios (June 25-26, 2026)
- Axios (July 8, 2026)
- TechCrunch (June 26, 2026)
- Malay Mail (April 11, 2026)
- Bloomberg via Incrypted (April 16, 2026)
- Ropes & Gray LLP analysis of AI Executive Order (June 3, 2026)
Disclaimer: The analysis above is based on publicly available data as of 2026-08-03. All statements attributed to named officials and companies are sourced from the media reports cited below. I am not affiliated with any of the companies mentioned unless explicitly stated. For the most current information, please visit the official sources linked throughout this article.
Limitations: This article covers events up to July 2026 and reflects reporting available at that time. Government actions and company releases evolve quickly; details may have changed since publication. Statements from unnamed officials are reported secondhand and cannot be independently confirmed.