Sam Altman Says OpenAI Has No IPO Timeline — Wants Safety Cases In Place Before Going Public

• Altman said OpenAI could consider an IPO after running “a few safety cases” and gaining a better understanding of how to safely contend with the next level of AI.
• He said the recently withheld OpenAI model was only “a little bit worse” on several evaluations and stressed there was no “big scary thing” behind the decision.
• Altman noted that AI safety cannot be treated solely as an engineering problem because model alignment remains a scientific challenge.

OpenAI CEO Sam Altman said Tuesday the company has no particular timeline for an initial public offering (IPO), indicating that safety work around the next generation of artificial intelligence (AI) remains a higher priority than going public.

“I don’t have a particular timeline in mind,” Altman said during an interview with CNBC at OpenAI’s Developer Day. An IPO could become more realistic once the company has run “a few safety cases” and feels it understands how to safely contend with the “next level of AI,” he said.

Altman said he does not want to pursue an IPO while conditions around AI development and safety remain in flux. “I really think this is a time to put safety and mission first,” he said.

Why OpenAI Held Back A Model

Asked about OpenAI's recent decision to hold back a new model, Altman said it fell short on some of the evaluations the company uses to gauge safety and alignment, rather than pointing to a specific serious incident.

The model was “a little bit worse on a few of the evals” OpenAI uses, Altman said, characterizing the decision as an “abundance of caution.” “There’s not a big scary thing in this case,” he said.

He described withholding or reworking models that do not meet OpenAI’s standards as part of the normal development process. 

More broadly, Altman said alignment, safety, monitoring and security need to remain “way ahead of capabilities” as models improve. He described that approach as “pacing” progress.

Safety Goes Beyond Engineering

Altman also pushed back on the idea that AI safety can be solved entirely through engineering. “We still have discovery about how to align these models,” he said, arguing that the industry faces a scientific problem alongside the engineering work required to make AI systems safer.

On liability, Altman said he expects a multi-level framework to emerge depending on whether a failure originates in the model, how a developer uses it, or intentional misuse by an end user. 

He compared the possible structure to the automotive industry, where responsibility can fall on different parties depending on what went wrong.

Altman also expressed cautious optimism about discussions in Washington and international coordination, while acknowledging that previous expectations for policy progress had not always materialized.

See Also: Apple CEO John Ternus Reportedly Weighs A Leaner Organization – Here’s The Key Things Under Consideration

For updates and corrections, email newsroom[at]stocktwits[dot]com.