Sam Altman: OpenAI won't go public until it can make confident safety claims

OpenAI CEO Sam Altman says his company will not go public until it can make stronger, verifiable promises about the safety of its models — and he offered no timeline for when that might happen.

"We intend to continue with AI progress … but as the models have had this surge forward in capability, and we see more of that ahead of us, we have got to be able to make confident safety claims," Altman said Tuesday, speaking to reporters in a Q&A session after his DevDay keynote.

He also warned against the opposite extreme. Waiting too long for an IPO, he said, would be "bad for the world."

The remarks address a question that has followed OpenAI for months: when will the company hold an initial public offering? Altman's answer makes clear that the answer is tied directly to the pace of model development — and to OpenAI's ability to certify that increasingly capable systems are safe.

Why an IPO depends on safety

The logic Altman laid out connects two things that are usually discussed separately: capital markets and model governance. As OpenAI's models gain capabilities in what he described as a "surge forward," the claims a public company must be able to stand behind become harder to make. A publicly listed company faces disclosure obligations and shareholder scrutiny. Making "confident safety claims" about frontier models — the kind that would satisfy investors and regulators — requires a level of assurance OpenAI does not yet believe it can provide.

At the same time, Altman framed indefinite delay as costly too. His comment that waiting too long would be "bad for the world" suggests he sees risks in keeping a company of OpenAI's scale and influence private forever, whether because of reduced transparency, constrained capital structures, or the broader stakes of AI development itself.

The result is a deliberate tension. OpenAI is telling the market: not yet, but not never.

A sector under scrutiny

Altman's comments do not come in a vacuum. They follow months of controversy over whether OpenAI and its rivals can be trusted to develop and deploy frontier AI systems responsibly — a debate that has drawn in researchers and watchdogs questioning the industry's safety practices.

That controversy frames why the IPO question matters beyond finance. OpenAI is one of the most valuable and closely watched AI companies in the world, and the terms on which it eventually enters public markets would set a precedent for how AI firms handle accountability, disclosure, and safety assurance. If safety certification becomes the gate for going public at OpenAI, it could shape expectations for competitors weighing their own paths to market.

The stakes cut both ways. Critics of the industry have argued that the race to build more capable models outpaces the ability to evaluate them. Altman's statement — that capability gains are surging and more are "ahead of us" — is, read one way, an acknowledgment of exactly that gap between what models can do and what their developers can promise about them.

What Altman actually said

Two claims anchor his position, and both are worth keeping verbatim.

First, on progress and safety: "We intend to continue with AI progress … but as the models have had this surge forward in capability, and we see more of that ahead of us, we have got to be able to make confident safety claims."

Second, on the cost of waiting: an IPO delayed too long would be "bad for the world."

Together, the statements position OpenAI's public-market debut as contingent on a milestone the company cannot yet date: the ability to make safety claims about its models with confidence. That is a notably different framing than the usual IPO speculation, which centers on valuations, revenue, and investor appetite. Altman has effectively made model assurance — not financial readiness — the gating condition.

No timeline, and that is the story

Altman declined to give a firm timeline for a public offering. In the current environment for AI companies, with intense competition among frontier developers and unresolved public debate over safety practices, an open-ended answer keeps OpenAI's options flexible. It avoids committing the company to a date it might miss or be forced to meet before its safety story is ready.

For investors, employees, and observers tracking the AI industry's maturation, the message is that OpenAI's path to public markets runs through its safety research — however long that takes.