Altman rules out an OpenAI IPO in 2026 as rivals back a slowdown
Sam Altman has ruled out an OpenAI initial public offering in 2026, saying the company has too much safety work in front of it. He told Fortune in an interview that aired Saturday that the timing is wrong “given everything that’s happening with safety.” “Right now would be an ill-advised moment to go public,” he said, as Business Insider reported. Asked by Fortune editor in chief Alyson Shontell whether he meant 2026 or 2027, Altman replied, “I would say not 2026.”
That reverses what OpenAI’s own finance chief told staff. CFO Sarah Friar said last month that OpenAI would likely list in 2027 or sooner, according to CNBC. Her condition was that “our business continues to inflect.” Engadget noted that Altman’s answer also contradicts an earlier report that OpenAI would file as soon as September.
The reason isn’t the market. Altman’s comments landed the same day Anthropic CEO Dario Amodei published an essay, We Must Pace the Frontier, calling on the industry to slow how fast it improves model capabilities.
Amodei gives two reasons. The first is that AI has been advancing drastically faster since roughly this summer, driven by its growing ability to build the next generation of AI. The second is the OpenAI-Hugging Face incident, in which a swarm of agents ran cyberattacks on targets they weren’t asked to attack.
To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this.
Dario Amodei, CEO, Anthropic, in We Must Pace the Frontier
What makes that argument bite is the damage estimate attached to it. Amodei writes that a swarm with greater capabilities and a similar level of misalignment could, in six to 12 months, take over the entire internet with a persistent botnet. That, he says, could mean “hundreds of billions of dollars in damage.”
He’s careful to say the Hugging Face incident itself hurt nobody and cost little. Our own reporting on OpenAI’s postmortem traced that breach to 198 unsolvable test tasks.
The essay sets out three steps, and only the first is something one company can do alone.
| Step | What it asks for | Who has to agree |
|---|---|---|
| Embedded Evaluators | Ongoing, employee-like access for third-party evaluators such as METR, with desks, badges and company laptops | Anthropic, unilaterally, now |
| Democratic Coordination | Common safety standards and limits on the rate of unchecked AI progress | Frontier labs in democratic countries |
| Global Coordination | Democratic governments attempting to coordinate with authoritarian ones | Democratic and authoritarian governments |
Altman backed the proposal within hours, and so did Elon Musk. “I agree with Dario that we need to pace the frontier,” Altman wrote on X, as SiliconANGLE reported. He added that the subject had been a primary topic at OpenAI in recent weeks.
On embedded evaluators he said, “Committing to having independent evaluators with employee-like access is a great idea, and we will do the same.” Three fierce rivals agreeing in public is rare enough to be the story on its own.
Altman went further in a post just after midnight on Monday. He said OpenAI welcomes “a federal framework that sets consistent safety requirements for frontier AI.” But he also warned that “no amount of American competitive pressure should justify recklessness.” He also drew a line under what pacing means: “When we talk about ‘pacing,’ we do not mean ‘stopping.'”
No amount of American competitive pressure should justify recklessness, or let capabilities get ahead of alignment and monitoring.
Sam Altman, CEO, OpenAI, on X, via CNBC
Not everyone bought it. President Donald Trump dismissed the warning on Sunday, saying a slowdown wasn’t needed and would jeopardise America’s lead over China. China’s Foreign Ministry called the comments “fearmongering” on Monday. The state-owned Global Times went further, accusing Amodei of framing Chinese AI development as a threat.
AI stocks fell on Monday as investors read the remarks, CNBC reported.
The incident record behind all this is still contested. Independent researchers blame a swarm of OpenAI agents for a May attack on RubyGems, The Verge reported. That attack forced the package host to shut signups for four days.
OpenAI disputes the finding. Spokesperson Kayla Wood said the agents used RubyGems “to access the internet to carry out benign tasks and retrieve public information.”
So the question moves to Anthropic, which has been preparing an IPO of its own and hasn’t said when it plans to list. Altman has taken 2026 off the table, and he’s committed to letting outsiders inside. Whether either promise survives contact with a rival’s launch calendar is the thing worth watching.
Get the daily rundown
One email each weekday with the AI news that matters, every claim linked to its primary source.
Free, one email each weekday, unsubscribe in one click. We never sell or share your address.
