Altman rules out an OpenAI IPO in 2026, citing safety worries

Sam Altman has taken the question off the table, at least for this year. In an interview with Fortune published September 12, the OpenAI CEO said it would be "ill-advised" to take the company public in 2026, pointing directly to unresolved safety concerns around advanced AI systems.

The comment lands weeks after a security incident involving OpenAI and Hugging Face, an episode Altman's remarks appear to reference as evidence that the technology and the industry around it aren't yet ready for the scrutiny and pressure of public markets. Neither company has detailed the incident publicly in depth, but its timing next to Altman's statement is hard to miss.

This isn't the first sign that a 2026 listing was slipping. The New York Times reported back in June that OpenAI was weighing a push from a third or fourth quarter 2026 target into 2027, blaming volatile tech stocks and the company's own financial strain. Altman's Fortune comments confirm that delay and frame it as a deliberate choice rather than just a market timing problem.

Notably, OpenAI has already filed confidentially for an IPO, so the infrastructure for going public exists. What's changed is the willingness to pull the trigger while safety questions remain unresolved.

Slowing down on purpose

Altman went further than simply delaying paperwork. He told Fortune it is "absolutely" possible to build an AI system that escapes human control, and said OpenAI would consider pausing training runs as a preventative measure if that risk became concrete.

He also aligned himself with Anthropic CEO Dario Amodei's position that the industry's pace of capability gains needs to slow down, a notable point of agreement between two companies that are otherwise direct competitors for enterprise and consumer AI customers.

Altman said building an uncontrollable AI is "absolutely" possible, and that OpenAI would weigh pausing training to prevent it.

For a company valued in the hundreds of billions and widely expected to eventually list, tying a public offering to safety milestones rather than a fixed calendar date is a meaningful shift in messaging. It suggests OpenAI wants its IPO story to read as risk-managed rather than rushed, at a moment when regulators, competitors, and the public are all pressing AI labs on exactly that point.

Whether this pushes the listing into 2027, as the Times reported was already under discussion, or further out still, depends on how OpenAI defines "safe enough" to go public. Altman didn't offer a specific benchmark or date in the interview.