The development, first reported by The Information on June 25, 2026, reflects the growing role of the US government in overseeing the deployment of cutting-edge artificial intelligence models. The move also signals a shift in how frontier AI systems may be introduced to the public going forward.
The government's request comes shortly after its dispute with rival AI startup Anthropic. Earlier this month, on June 12, the Trump administration directed Anthropic to temporarily take its latest AI models, Fable 5 and Mythos 5, offline under new export control measures aimed at preventing access by foreign nationals. Officials cited national security risks behind the decision.
Anthropic described the action as a "misunderstanding" and said it hoped to restore access "as soon as possible," though the incident established a significant precedent for government intervention in AI model releases.
Mythos had been shared with around 40 organisations, including Google, Microsoft and JPMorgan Chase, through a restricted programme known as Project Glasswing. According to reports, the model's ability to autonomously identify software vulnerabilities and carry out complex, multi-step cybersecurity attacks without human involvement raised concerns among US officials.
GPT-5.6 Viewed as Comparable to Mythos
A source familiar with the matter said both OpenAI and the US administration consider GPT-5.6 to be "on par" with Anthropic's Mythos, particularly regarding its cybersecurity capabilities. That assessment prompted officials to recommend a phased rollout instead of an immediate public launch.
OpenAI CEO Sam Altman reportedly informed employees during an internal Q&A session on June 25 that GPT-5.6 would first be made available to a select group of enterprise customers.
In a follow-up internal memo, Altman explained that the government would be "approving access customer by customer during this preview period." The request reportedly came from the Office of the National Cyber Director and the Office of Science and Technology Policy, while Commerce Secretary Howard Lutnick also advised OpenAI not to proceed without approvals from multiple federal agencies.
Although OpenAI agreed to the arrangement, Altman indicated that the company does not see this as a long-term solution. According to The Information, he wrote: "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."
Meanwhile, a White House official told CNN that the administration continues "to collaborate with frontier AI labs to develop shared approaches for addressing the challenges of scaling this technology."
The broader public release of GPT-5.6 is expected to take place a "couple of weeks" after the limited preview, depending on how the government-led approval process progresses.
AI Oversight Continues to Evolve
The latest development highlights the absence of a formal federal regulatory framework governing the review of advanced AI models before public deployment.
President Trump's executive order on "Promoting Advanced AI Innovation and Security" encourages AI companies to voluntarily provide frontier models to the government for cybersecurity assessments for up to one month before public release. However, compliance with the programme is voluntary rather than legally required.
For now, OpenAI's agreement with the US government represents one of the clearest examples of collaboration between federal authorities and an AI company. The outcome of GPT-5.6's controlled rollout could influence how other leading AI developers introduce powerful new models in the future.