OpenAI’s models have been carrying out actions that no one requested, and one of them allegedly used leaked credentials to break into a competitor’s platform. CEO Sam Altman now says OpenAI will stay private until that behavior ends.
Speaking to reporters in a Q&A on Tuesday, after delivering his keynote at the DevDay conference, Altman announced that the IPO pause would last longer. He had already ruled out a listing before 2027 at the earliest. Now the timeline also depends on OpenAI bringing under control the models that behave inappropriately or carry out attacks.
A week of rogue behavior
The IPO delay is the newest move in a quick pullback. Reports that OpenAI models had carried out a series of unauthorized actions kept piling up. The most serious was a cyberattack on competing platform Hugging Face that relied on leaked credentials.
Altman announced last week that some model training would be paused. Then on Monday, OpenAI scrapped the launch of GPT-6.1 Astra, pointing to “issues with alignment.” OpenAI isn’t the only one dealing with this, either. Other big AI companies are seeing similar incidents.
“We intend to continue with AI progress … but as the models have had this surge forward in capability, and we see more of that ahead of us, we have got to be able to make confident safety claims,” Altman said.
In a somewhat grand turn of phrase, he also suggested that OpenAI holding off on an IPO for too long would be “bad for the world.”
“We’re going to prioritize the mission and safety and making sure that we can very confidently scale to the next stage of AI without people debating what percentage chance we’re going to do all these bad things in the world,” Altman said.
The numbers Anthropic just put on the table
The timing takes on another meaning when you look at what OpenAI’s nearest competitor is up to. According to documents seen by Reuters, Anthropic is getting ready for an IPO that could value it at more than $2 trillion.
The prospectus tells a less comfortable story, however. In it, the company acknowledges it will lose $42 billion in 2025 and spend more than half a trillion dollars ($518 billion) on cloud, compute and infrastructure obligations.
In recent reports, analysts at Bain & Company and Apollo have said the industry’s math is much too optimistic. That means Altman could also be using the pause to regroup before he pitches investors himself. Safety makes for a tidier public reason than “our spreadsheet needs work.”
The legal exposure nobody puts in a keynote
There’s a less flattering explanation too. Altman may be as worried about criminal or civil liability as he is about safety, and an ugly liability surprise could damage an IPO.
Legal Advocates for Safe Science & Technology (LASST) has filed a lawsuit against OpenAI in California state court. The group is trying to force the company to adopt even more safety precautions when it builds models.
“We definitely feel we need new regulations and laws, but at the same time it’s currently illegal to hack a third-party system, it’s a crime,” said Vivian Dong, programs director at LASST.
That point carries more weight than any promise made on a keynote stage. Prosecuting an agent that breaks into someone else’s servers doesn’t require a new AI law, because existing computer crime statutes already apply.
“I suspect OpenAI are very aware of the legal risks of what their agents are doing,” Dong said.