OpenAI has agreed to delay the public launch of its newest AI model, GPT-5.6, after the White House requested a limited, staggered rollout due to concerns over the model's advanced capabilities. Instead of a typical public release, the model will initially be made available only to a small group of government-approved partners, with access approved on a customer-by-customer basis during a preview period.
In a staff memo reported by The Information, OpenAI CEO Sam Altman said the limited release was not the company's "preferred long term model" for launches, but added that OpenAI hoped to work with the government toward "a more sustainable approach for future releases." Altman said the broader, general release of GPT-5.6 is expected to follow a "couple of weeks later" if the initial preview phase goes smoothly.
Why the White House Got Involved
The request reportedly came from the White House's Office of the National Cyber Director and the Office of Science and Technology Policy, as the administration works to build a formal framework for testing and evaluating the security of powerful new AI models. According to sources, OpenAI and the administration view GPT-5.6's capabilities as being "on par" with Anthropic's advanced Mythos models, which raised significant concern in Washington over potential cybersecurity risks.
This is not the first time OpenAI has taken a cautious approach with a highly capable model — the company previously released its most cyber-capable system, GPT-5.5-Cyber, to a similarly limited group of organizations. The current request follows President Trump's executive order, signed earlier this month, which asks AI companies to voluntarily submit advanced models for a 30-day government review period before public release, though a formal framework for that process has not yet been finalized.
Echoes of the Anthropic Situation
The move comes shortly after the Commerce Department issued an export control directive that led Anthropic to pull its own frontier models, Claude Mythos 5 and Claude Fable 5, just days after their public release, citing national security concerns. While the OpenAI request came from the White House rather than Commerce, the back-to-back developments highlight growing federal scrutiny over how the most powerful AI systems are deployed.
In a public statement, OpenAI acknowledged the limitations of this approach, writing: "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." For continued coverage on the evolving regulatory landscape around frontier AI models, readers can follow ongoing reporting from Axios.
What Comes Next for AI Regulation
Industry experts have warned that the current approach to AI oversight remains inconsistent and lacks transparency. Brad Carson, head of the pro-AI safety super PAC Public First, described the existing framework as "an ad hoc, personalized, opaque, possibly lawless approach," arguing that any government intervention in AI releases needs to be paired with clear, consistent regulation. As the White House continues developing its evaluation framework over the coming weeks, both OpenAI and the broader AI industry are likely to face continued uncertainty over how future frontier model releases will be handled.