.png?w=1024)
US President Donald Trump’s administration has asked OpenAI to stagger the release of its next model, GPT 5.6, to a select group of partners.
Speaking during a staff Q&A session, OpenAI CEO Sam Altman said the company would launch GPT 5.6 in a limited preview, similar to Anthropic’s Mythos and GPT-5.5-Cyber. He said the launch sequence was the “best path” for getting the model released globally as soon as possible.
OpenAI received requests for a staggered release from two government agencies: the National Cyber Director and the Office of Science and Technology Policy, according to a report from The Information. Commerce Secretary Howard Lutnick also warned OpenAI not to launch the model without approval from other government agencies.
US government tightening its control of AI releases
The move marks a tightening of US government control over frontier model releases, following the debacle involving Anthropic and the release of its Fable model. Fable, which was built using frameworks similar to Mythos but with more guardrails, launched on June 8, but within a week, Washington had imposed restrictions on foreign access to the model.
The Trump Administration has been meeting with frontier AI model developers to discuss the best way to launch these models and quickly receive approval for foreign release.
There still does not appear to be much clarity on the approval process, with the current system a mish-mash of agency licensing requirements and informal approval channels. The administration plans to launch a voluntary approval system for AI models, but that is undercut by the fact that Washington can still ban foreign access or push a model into limited preview.
OpenAI was more open than its rival, Anthropic, with its Cyber release, providing earlier access to European governments, banks, and critical infrastructure suppliers. The allure of Mythos kept many European and other foreign governments and companies pushing for access, which may have contributed to its restriction.
Altman has said that Anthropic’s messaging around Mythos and Fable was ultimately what led to its restriction, calling it “fear-based marketing.”
Government activity may delay IPOs
The US government’s growing involvement in frontier AI releases may force both Anthropic and OpenAI to reevaluate their IPO plans. OpenAI filed confidentially in early June, a few days after Anthropic. Both are seeking valuations in the $1 trillion range.
OpenAI was the first to blink, with internal conversations already taking place about delaying its IPO until 2027. That would give Anthropic a first-mover advantage, but given its own battle with the White House, it may not find much value in going public earlier. The first week of SpaceX's public market debut may also have dampened expectations, as it is already trading below its initial $160 price.
What this means for businesses and AI users
A staggered GPT-5.6 release could mean some companies get access to OpenAI’s newest capabilities weeks or months before others. That may matter for enterprises testing advanced AI tools for cybersecurity, software development, research, financial services, or critical infrastructure.
For readers, the bigger takeaway is that access to frontier AI may become less predictable. New models may no longer arrive through broad public launches, especially if US officials believe a system could raise national security, economic, or geopolitical concerns.
That could create a split market where approved partners move faster, while smaller businesses, international users, and developers wait for wider availability.
Also read: OpenAI’s custom Jalapeño AI chip with Broadcom could help the company lower inference costs and reduce its reliance on Nvidia.