OpenAI had confidentially filed for an IPO in June, but had not committed to a timetable—a setup that kept alive expectations of one of Silicon Valley’s biggest market debuts. That prospect now appears firmly deferred. ¹

In a Fortune interview released Saturday, CEO Sam Altman said an offering this year would be the wrong priority as labs confront the risks posed by ever more capable AI. “Right now would be an ill-advised moment to go public,” he said. Pressed on whether that ruled out 2026, Altman replied: “I would say not 2026. Yeah, we got a lot of stuff to do.” ¹

His argument is not simply that OpenAI needs more time to build products. It is that the company needs latitude to put the brakes on. Altman said OpenAI had discussed pausing training runs when models reach new capability levels, allowing researchers to advance safety and alignment work before pressing ahead. “We will do more of that going forward,” he said. ¹

The warning is unusually stark coming from the head of the company driving much of the AI boom. Altman said it was “absolutely” possible to build an AI beyond human control, while pledging that OpenAI would act to prevent it—even if that meant stopping training. “There are risks we should not be able to incur on behalf of humanity,” he said. ²

That puts OpenAI closer to other frontier-lab voices urging restraint, including Anthropic chief Dario Amodei, than to a simple race-for-scale narrative. Altman said the company should not train systems unless it can make a credible safety case for their controllability and alignment. ¹

For investors, employees and San Francisco, the practical consequence is delay: a potential windfall remains private while OpenAI tests whether its mission can coexist with models powerful enough to challenge human oversight. Altman’s bottom line was clear: OpenAI must retain the ability to make choices “not obviously in the interest of our business and our shareholders.” ¹