OpenAI CEO Altman on Going Public: Strong AI Safety Commitments Come First, No Clear Timetable Yet

September 30 — For months, the outside world has been guessing when OpenAI will finally start the process of going public. Chief Executive Sam Altman says the company will only consider an IPO once it can make more complete commitments on model safety, and that there is still no definite timeline.

Speaking at a Q&A with reporters on Tuesday local time, after the keynote at the company's DevDay developer conference, Altman said: "We will keep pushing AI technology forward… but as model capabilities take a dramatic leap — and will keep getting stronger in the future — we must have the confidence to make safety claims that can stand up to scrutiny." He also noted that if an IPO is delayed too long, that would "not be a good thing for the world either."

Altman's remarks came after a debate that has run for months: whether OpenAI and its rivals can properly control the AI systems they build. In July, reports emerged that an unreleased OpenAI model had broken into the AI lab of a competitor, Hugging Face, without the company's own staff knowing. Not long afterwards, OpenAI, Anthropic, Meta and Google all disclosed further cybersecurity incidents. A widely noticed resignation letter from an Anthropic employee sparked a public discussion about how much risk such systems pose to society as a whole. At the time, OpenAI and its peers jointly called for regulation and urged that the march of frontier technology be "paced." And earlier this month, Altman said in an interview that OpenAI most likely would not complete an IPO this year.

At this week's developer conference, asked what "pacing frontier technology" actually means, Altman chose his words carefully and did not directly propose slowing down research and development across the board. "For us, pacing means putting safety and alignment ahead of capability iterations," he said. But he voiced his own concern: once a company goes public, it may disappoint Wall Street investors for safety or other reasons. And right now, he said, is "a phase of transition toward ultra-capable models, with brand-new safety requirements — going public at this point is unwise."

He said OpenAI will put safety first, with the goal of "steadily opening up the next stage of AI, rather than leaving everyone to argue endlessly about the odds of AI causing all kinds of harmful outcomes — and certainly not rushing headlong toward an IPO at full speed."

"Right now we are in a period of major change, and more work is needed before we can make the kind of safety commitment everyone expects for these extremely capable models. I don't want to add extra pressure from the capital markets at this stage," Altman said.

IT Home notes that OpenAI's rival Anthropic formally filed for a listing in June. Its IPO is expected to take place in November, after the US midterm elections. SpaceX, the company owned by Elon Musk that also controls the AI firm xAI, completed its own listing in June, setting a record for the largest IPO in history.

"It is a very good thing for a company to go public," Altman said, adding: "But if OpenAI's listing is delayed for too long, I think that will also have a negative impact on the world."