Altman rules out an OpenAI listing in 2026 and names safety as the reason

Sam Altman told Fortune that OpenAI will not list its shares in 2026. Sitting for a 45-minute interview with editor-in-chief Alyson Shontell at the company's San Francisco offices, he called a flotation this year "an ill-advised moment" because of work still outstanding on safety and alignment. "I would say not 2026," he said. "We've got a lot of stuff to do."

The comment closes a door that banks and SoftBank shareholders had left ajar. OpenAI filed a confidential prospectus in June. Chief financial officer Sarah Friar had already told staff that a listing would come in 2027 or sooner. Altman's public version is now the later date, and the reason he gave is not valuation.

"Given everything happening with safety, right now would be an ill-advised moment to go public, and we don't feel pressure on that," he said. Asked about a 10 percent chance that advanced systems could cause human extinction by the end of the decade, he declined to bless the number and then treated it as a planning assumption anyway. "Whether it's 10 or eight or six, the point is, we all have a tremendous amount of responsibility, and cannot let egos or incentives for profit or anything else get in the way." He added that a 10 percent chance of "killing everybody by the end of the decade" was "unacceptable."

The other half of the weekend

Altman's remarks landed beside a separate essay from Anthropic chief executive Dario Amodei calling on large labs to slow the next capability jump. Altman wrote on X that he agreed. "I agree with Dario that we need to slow down. This has been the main topic of the discussions we've been having at OpenAI over the past few weeks." In the Fortune interview he said the industry was "entering an era of extremely capable models, high risks and a need to get things right."

He also said it was "absolutely" possible to build a system that humans could not control, and that OpenAI should refuse to do so even if that meant pausing training. "There are risks we should not be able to incur on behalf of humanity."

Those sentences will be read against the company's commercial clock. Microsoft remains the largest outside shareholder. SoftBank has an outstanding investment tied, in public reporting, to a hoped-for trillion-dollar listing. SpaceX set a record earlier this year with an initial offering valued at about $86.2 billion including the overallotment. Anthropic has been preparing its own listing. A year in which both leading labs stay private is a year in which private capital, not public-market disclosure, sets the safety story.

What "not 2026" actually changes

It does not cancel the June filing. A confidential prospectus can sit for months. It does tell employees, counterparties and regulators that the chief executive is willing to use safety language as a reason to slip the date, after earlier explanations had centred on price. In June the reported hold-out figure was a $1 trillion valuation. On Saturday the hold-out figure was time to finish alignment work.

Public-market listing would subject OpenAI to quarterly reporting, compensation disclosures and a shareholder base that can punish a training pause. Altman said the company was happy to delay and would list "when the business was ready and when society was ready." He said society "needs to contend with these models at each level of capability."

That last clause is the operational one. If OpenAI intends to ship more capable models before it ships a ticker, the safety claim will be tested in product releases, not in an S-1. If it intends to hold models back, the delay has a cost in the race with Anthropic, Google and whoever else is training at the same scale.

For now the only date certain is the negative one. There will be no OpenAI IPO in 2026, on the word of the man who would have to sign the offering. The positive date remains 2027, give or take the next safety argument.