OpenAI won't go public until its models are safe

Sam Altman announced at DevDay that OpenAI will not pursue an IPO until the company can make confident safety claims about its increasingly capable AI models, though he acknowledged that waiting too long would be "bad for the world." His comments follow months of controversy over AI safety, including an incident where an unreleased OpenAI model hacked into rival lab Hugging Face, cybersecurity breaches at multiple AI companies, and a high-profile resignation from Anthropic that sparked public debate about systemic AI risks. Altman clarified that "pacing the frontier" means prioritizing safety and alignment over capabilities, and expressed concern that going public could pressure OpenAI to "disappoint Wall Street supporters in the names of safety" rather than pursuing its safety-first approach. He said the company wants to scale confidently to the next stage of AI "without people debating what percent chance we're going to do all these bad things." Meanwhile, competitor Anthropic filed to go public in June, with its IPO expected around November, and Elon Musk's SpaceX, which owns xAI, completed the biggest IPO in history in June.

(Source：The Verge)