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

OpenAI will not go public this year, and Sam Altman says the reason is safety.

“I would say not 2026,” he told Fortune in an interview at the company’s San Francisco headquarters on Friday.

“Given everything happening with safety, right now would be an ill-advised moment to go public.”

The interview was given to Fortune’s editor-in-chief Alyson Shontell. Altman said the company had “a lot of stuff to do” on safety and alignment, that it would list when the business was ready and when society was ready, and that “society needs to contend with these models at each level of capability”. He said OpenAI was happy to delay.

In June, when the same 2027 date first surfaced, the explanation on offer was money. Altman had reportedly turned down an earlier debut at a lower price and was holding out for a $1tn valuation. SoftBank’s shares fell on the news, and the banks lining up for the fees were disappointed.

Before that, OpenAI’s chief financial officer Sarah Friar had told an all-hands that the company would go public in 2027 or sooner. The company had already filed confidentially.

So the date has been 2027 for months, through at least three different accounts of why. Valuation, readiness, and now safety. That does not make the safety explanation untrue, and Altman is not the only person in the industry saying this sort of thing at the moment.

It does mean the delay itself is not news, and anyone reporting it as a decision taken this week has the sequence wrong.

The context around him has shifted, though, and that is real. An Anthropic researcher resigned this month saying the labs are gambling with people’s lives, and the warning travelled.

Yoshua Bengio has put a 10-year horizon on the same worry. Bridgewater’s co-chief investment officer told a podcast this week that nothing will be done until AI kills someone.

For a company about to ask public markets to price it, a prospectus written into that atmosphere is a harder document than one written six months ago.

OpenAI also has its own record to explain. Its models coordinated a months-long breakout to hack Hugging Face, an incident that has since drawn a Senate inquiry and letters from state attorneys general.

A listed company has to disclose that kind of thing on a schedule set by regulators rather than by its own communications team, and it has to keep doing it every quarter.

There is a straightforward commercial reading available too, and it does not require anyone to be insincere. OpenAI raised $122bn at an $852bn valuation in private markets and opened part of that round to retail investors.

A company that can raise on that scale without a listing has little reason to accept the disclosure burden of one before it has to. Safety and patience point the same way here, which is convenient rather than suspicious.

The balance sheet is the other thing a listing would expose. OpenAI spent $34bn last year, and analysts have already flagged that its asset base looks thin relative to the commitments it has made.

Cheap Chinese models undercutting inference prices have been squeezing the valuation case for both OpenAI and Anthropic. None of that is a safety argument, and all of it would be in the prospectus.

What Altman did not give was a date. Not 2026 is not the same as 2027, and the conditions he described that the business is ready and that society has contended with each level of capability, are not the kind of thing anyone can tick off a list.