OpenAI to Restrict GPT-5.6 Release at U.S. Government’s Request OpenAI’s launch of its most powerful AI yet, GPT-5.6, has turned into a live test of how far the U.S. government can go in steering private AI deployments, pitting security concerns against fears of a creeping licensing regime.

On June 2, President Trump signed an AI security executive order asking leading firms to “voluntarily” submit advanced models for pre‑release review, stopping short of formal licensing but creating a new channel for federal scrutiny. That framework gained teeth after Commerce ordered rival Anthropic to suspend access to its Mythos 5 and Fable 5 models over export‑control and jailbreak worries, effectively forcing them offline.

In the weeks that followed, the White House’s Office of the National Cyber Director and Office of Science and Technology Policy began working with OpenAI on its upcoming GPT‑5.6 release, previewing the model’s “Mythos‑like” capabilities and drafting a security evaluation process. Multiple agencies, including Treasury and Commerce, then asked OpenAI to “stagger” distribution so users could be vetted first.

News of that request broke on June 25. TechCrunch reported that “The White House is asking OpenAI to slow roll the release of its new model over safety concerns,” with officials planning to approve access “customer by customer” during an initial preview. Axios described it as the first time Washington has preemptively asked a U.S. AI company to restrict a launch, while The Verge noted that the government would sign off on enterprise customers “on a case-by-case basis.”

Less than 24 hours later, on June 26, OpenAI went ahead with GPT‑5.6 — but only in a constrained form. The company announced a “limited preview” of Sol, Terra, and Luna, emphasizing that the models are “especially skilled at coding, cybersecurity, and biology” and backed by its “most robust safety stack to date.” An accompanying system card stressed that safeguards were designed “to deliver these models safely and at scale, around the world.”

In practice, however, access is tightly controlled. Axios reported that GPT‑5.6 is available initially to “around 20 companies” whose participation has been approved by the government, and The Next Web said Sol went to “roughly 20 partners whose names were individually approved by the US government.” The Financial Times similarly described a release to “select users vetted by US government” under a limited preview.

OpenAI frames the constraints as temporary and tied to national security. Business Insider quotes the company saying it is starting with “a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly,” in line with an agreement that lets the Pentagon use its models. In a blog post cited by Axios, OpenAI warns that the current approval regime “keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them,” and adds: “We don’t believe this kind of government access process should become the long-term default.”

From the administration’s perspective, GPT‑5.6 is precisely the kind of frontier system that justifies early intervention. Axios reports that officials stepped in because the model has “Mythos-like” capability and argued that “models of that caliber” must demonstrate adequate safeguards before broad deployment. The Financial Times says agencies requested a staggered rollout specifically “to vet users,” signaling a product‑review mindset more familiar from defense contracting than consumer software.

Industry reaction is mixed. Some technologists see continuity with past caution around powerful models: a viral post from Hugging Face CEO Clement Delangue resurfaced that “Dario once delayed the release of GPT-2 back at OpenAI, claiming it was too dangerous,” suggesting that security‑driven slowdowns are not entirely new. Others, like former White House AI adviser Dean Ball, argue Trump’s order has created a “de facto involuntary licensing regime” for frontier AI, warning in TechCrunch coverage that vague safety standards and case‑by‑case approvals could cause “endless launch delays” and hand an advantage to China.

OpenAI itself is trying to walk a narrow line between cooperation and pushback. Axios notes that the company has been “previewing GPT‑5.6 with the government for the past month” and sees acquiescing now as “the strongest path to broader availability in the coming weeks,” while it works with the administration on a repeatable process for future models. Yet TechCrunch emphasizes that the firm is “not happy with the arrangement” and explicitly rejects making government‑managed access the norm.

For developers and enterprises outside the chosen 20, GPT‑5.6’s debut is less a product launch than a signal: in the U.S., the frontier of commercial AI now effectively runs through Washington.

Continue reading https://foxvector.com/stories/019f082e-fdaa-1309-732d-2f3bcfaf3451