The White House is asking OpenAI to slow roll the release of its new model over safety concerns
TL;DR
OpenAI reportedly will not release GPT 5.6 broadly at first. Instead, it plans to give access only to selected partners after pressure from the Trump administration. Sam Altman reportedly told staff that the government would approve access customer by customer during the preview. A broader release could follow a few weeks later if the limited rollout goes well.
Nauti's Take
The safety argument is plausible, but the communication is thin. OpenAI, Anthropic, and the government all benefit when the newest models look dangerous, desirable, and exclusive at the same time.
That is why more transparency matters: what capabilities were tested, what thresholds apply, and who decides access? Without those answers, the slow rollout sits somewhere between safety measure and power play.
Briefingshow
This is more than a standard staged product launch. If a frontier model ships only after government review and customer-by-customer approval, control shifts from product teams and markets toward public oversight bodies. At the same time, the actual risk level of GPT 5.6 remains hard to judge because the technical details are not public.
Sources