The White House is now hand-picking who gets OpenAI GPT-5.6
Sam Altman told OpenAI employees on Wednesday that the Trump administration asked the company to stagger the release of GPT-5.6. Not delay it. Not block it. Stagger it. The model ships first to 20 government-approved partners, routed through Amazon Bedrock, and then the government decides who else gets in.
Altman’s exact words to staff, per a memo obtained by The Information: the government will be “approving access customer by customer” until the preview period ends. A wider release might follow “within weeks.” Might.
This is the second time in June the White House has inserted itself between a frontier AI lab and its customers. Two weeks ago, the Commerce Department forced Anthropic to pull Fable 5 and Mythos 5 from the market entirely. Now OpenAI gets a softer version of the same treatment: you can ship, but we choose the audience.
The White House justified the restriction by telling OpenAI that GPT-5.6 has capabilities comparable to Anthropic’s Mythos model, according to reporting from The News International. That is the same Mythos that the Commerce Department deemed too dangerous for public access.
Trump signed an executive order last month requiring AI companies to submit frontier models for a 30-day government review before public release. The problem: the review mechanism does not exist yet. There is no framework, no published criteria, no appeals process. It is just officials making calls about who can use what software.
Altman did not hide his frustration. In the staff memo he wrote: “We’ve made clear to the US 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.” Translation: we are complying because we have no choice, not because we agree.
Brad Carson, head of Public First, a bipartisan pro-AI-safety political action committee, put it more bluntly to CNN. He called the current approach “ad hoc, personalised, opaque, possibly lawless.” He added that recalling dangerous products is legitimate, “but it has to be done in a way consistent with transparency and basic fairness.” Right now it is none of those things.
The stakes are not abstract. OpenAI spent roughly $14 billion serving ChatGPT in 2025 on third-party GPUs. Its custom Jalapeño inference chip, unveiled with Broadcom on June 24, is supposed to cut that cost in half. But none of that matters if the government controls who can buy the product.
Prediction markets have already moved. Polymarket odds of GPT-5.6 shipping by end of June collapsed to near zero after the news broke. Crypto Briefing noted that the staggered release aligns with the executive order’s 30-day voluntary review window, and that Meta and Google face the same pressure.
A White House official told reporters the administration is “collaborating with frontier AI labs to develop shared approaches for addressing the challenges of scaling this technology.” Shared approaches. That is a polite way of saying the government now decides what gets shipped, to whom, and when.
Two months ago, the question was whether AI models were too dangerous to release. Now the question is whether the US government has the authority to ration them without a law saying it can.