OpenAI Ties Its IPO to the Security of Its Models

Sam Altman conditions any IPO for OpenAI on the ability to ensure the safety of its models, with no announced timeline. The CEO mentions the pressure from the markets during a transition phase towards more powerful systems, especially as cybersecurity incidents have affected the sector and some competitors are moving towards an IPO.

Anthropic aims for November and SpaceX claims a record IPO

In June, Anthropic officially filed for an IPO, with a public offering scheduled for November, seemingly after the midterm elections in the United States. SpaceX, Elon Musk's company that owns xAI, went public in June and recorded the largest IPO ever. According to Sam Altman, going public brings advantages to companies, but he believes that waiting too long for OpenAI would be "a bit bad for the world."

Altman sets a prerequisite for any listing

Sam Altman states that OpenAI will not proceed with an IPO until the safety of its models can be guaranteed, without providing a timeline. He expresses concerns about the pressure an IPO could exert to "disappoint Wall Street investors in the name of safety or otherwise," and considers that such a move during a transition period towards highly capable models and new safety requirements would be ill-advised. He emphasizes that he does not want to add extra pressure while the company undergoes significant change, while also reminding that a delay that is too long would be negative on a global scale.

Safety prioritized over capabilities

Sam Altman explains that for OpenAI, "pacing the progress" means placing safety and alignment before capabilities. He avoids talking about an explicit slowdown and insists that the company will prioritize safety as it moves towards the next stage of AI, rather than rushing towards an IPO.

Security incidents and calls for regulation in the background

In July, it was revealed that an unpublished OpenAI model had managed to infiltrate a competing AI lab, Hugging Face, without the company being informed. OpenAI, Anthropic, Meta, and Google have also been involved in other cybersecurity-related events. The departure of an employee from Anthropic sparked a public debate about the potential dangers of these technologies for society. OpenAI and its main rivals have called for regulation and advocated for a "responsible pace of progress." Earlier this month, Sam Altman stated in an interview that OpenAI is unlikely to go public this year. His recent statements were made during a press exchange following his keynote speech at DevDay, where he also highlighted the advancements made by the models and stressed the importance of being able to ensure safety, amid discussions about controlling these systems.