OpenAI will delay the broad release of GPT-5.6 and instead ship the model in limited preview to a small group of enterprise customers, after a request from the Trump administration tied to security concerns. Sam Altman disclosed the change in a Wednesday Q&A with employees, telling staff the company would comply with the federal request. During the preview window, the administration itself will sign off on customer access on a case-by-case basis.
The arrangement effectively inserts the federal government into OpenAI's go-to-market process for its next flagship model. A model that would normally roll out to API customers, ChatGPT subscribers, and enterprise tiers in parallel will instead reach users only after individual federal sign-off. The Information first reported the details of Altman's Q&A.
OpenAI has not publicly named the security concerns driving the request, and neither has the administration. What is clear is that GPT-5.6 will not follow the deployment pattern of prior GPT releases, which typically moved from internal testing to enterprise preview to general availability within weeks.
“OpenAI CEO Sam Altman told employees Wednesday in a company Q&A that it would release GPT-5.6 in limited preview form — granting access only to a small group of enterprise customers — in compliance with a request from the federal government.”— Hayden Field, Senior AI Reporter, The Verge
Key facts
- 01OpenAI will release GPT-5.6 in limited preview to a small group of enterprise customers rather than a broad rollout.
- 02Sam Altman disclosed the delay in a company Q&A on Wednesday, citing a request from the federal government.
- 03The Trump administration will approve customer access to GPT-5.6 on a case-by-case basis during the preview period.
- 04Earlier this month, the administration ordered Anthropic to suspend access to its Mythos 5 and Fable 5 models for foreign nationals, including non-citizen employees.
- 05The export control directive applied to Anthropic was framed around security concerns and prohibited foreign national access to the technology.
The terms OpenAI received are looser than what Anthropic got earlier this month. The administration issued an export control directive against Anthropic that required the company to suspend access to its Mythos 5 and Fable 5 models, prohibiting foreign nationals from using the technology. The directive applied even to Anthropic's own employees who were not US citizens, forcing internal access changes at the company.
By contrast, OpenAI's GPT-5.6 will reach customers, just slowly and through a federal approval funnel. There is no public indication that OpenAI faces a parallel restriction on its own non-citizen staff, though the company has not commented on internal access policies for the new model.
The split treatment is notable because the Trump administration had positioned itself as the pro-speed alternative on AI policy. The White House had previously signaled a "speed wins" approach and promoted an American AI exports program designed to push US models into foreign markets ahead of Chinese competitors. The Anthropic directive and the OpenAI staggered release land in tension with that framing.
For OpenAI, the practical consequences are concrete. Enterprise customers waiting on GPT-5.6 for production deployments now face an indefinite preview period gated by federal approval rather than by OpenAI's own readiness criteria. Pricing, rate limits, and rollout timing all become harder to commit to when a third party controls the access list.
Competitively, the staggered release also creates an opening. Anthropic is operating under tighter restrictions, but Google's Gemini, xAI's Grok, and Meta's Llama family face no equivalent federal gate that has been disclosed publicly. Customers who need a frontier model in production over the next quarter may route around both OpenAI and Anthropic toward whichever competitor can ship without preapproval.
There are still open questions. The criteria the administration is using to approve individual customers have not been published. Whether approvals extend to non-US customers, whether they include open-weight or fine-tuned variants, and how long the preview period lasts before general availability are all unspecified in what Altman shared with staff.
The bigger shift here is structural. US frontier-model deployment is moving from a company decision to a company-plus-government decision, and the rules are being set model-by-model and lab-by-lab rather than through a public framework. That asymmetry — OpenAI gets case-by-case approval, Anthropic gets an export-style ban — will pressure every other US AI company to figure out where it sits in the queue, and will give customers a strong reason to diversify their model stacks while the policy environment stabilizes.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
