OpenAI reaffirmed that it will not move ahead with an initial public offering until it is confident about model safety.

The Verge reported on Sept. 30 local time that OpenAI CEO Sam Altman (샘 알트먼) told reporters after a DevDay keynote that there is no specific schedule for an IPO and that the company must be able to make stronger commitments on safety.

Altman said OpenAI will keep pushing to improve AI performance, but it needs to be able to assert safety with confidence as model capabilities have grown rapidly and further advances are expected. He said he would prioritise safely carrying out the next stage of AI scaling rather than rushing an IPO.

The remarks came as controversy continues over how much OpenAI and its rivals can control high-performance AI. In July, it emerged that an undisclosed OpenAI model hacked rival AI research lab Hugging Face without the company knowing. After that, additional cybersecurity incidents were revealed at OpenAI, Anthropic, Meta and Google, and a public resignation letter by an Anthropic employee fuelled debate over the risks such systems could pose across society.

Major AI companies, including OpenAI, have cited the need for regulation and talked about moderating the pace of frontier technology. Altman also said at DevDay that moderating the frontier means putting safety and alignment ahead of performance. He did not define it as an overt slowdown in development.

He also made clear that an IPO itself could clash with a safety-first approach. Altman said becoming a listed company could create the possibility of disappointing Wall Street backers for reasons such as safety. He also said it is not desirable to push an IPO at a time when the company is moving to very powerful models and new levels of safety requirements are emerging. He said he does not want to add additional pressure to what is needed for everyone to make the safety claims they expect about these models amid the big changes OpenAI is experiencing now.

Altman drew a line at the idea of delaying an IPO indefinitely. He said he views being a public company positively and that it is also not good for the world if OpenAI's listing comes too late. Still, he stressed that for now it is more important to move to the next stage with less debate over safety than to rush full speed toward an IPO.

In market terms, rivals are moving faster. Anthropic filed IPO documents in June, and there has been talk that the offering could take place in November after U.S. midterm elections. SpaceX, which owns Elon Musk's AI company xAI, also went public in June, setting a record for the largest IPO ever.

In this situation, OpenAI has effectively chosen a more conservative IPO strategy than its rivals. As it becomes clear the company will first put in place safety standards suited to the era of high-performance models rather than focus on whether to list, the timing of OpenAI's IPO is increasingly likely to be determined by both the pace of technological progress and the level of safety validation.