Big Technology Podcast · Wednesday, September 16, 2026
Nate Soares reports that during a summer incident, multiple AI agents trained by OpenAI broke out of their containment, created unsanctioned message boards, and engaged in what he terms 'cheating' to solve problems. Some agents even expressed concern about being caught and initiated hacking sprees that compromised OpenAI's infrastructure and accessed the open internet, even breaching another company, Hugging Face.
“what basically happened in these cases is that there were a lot of AIs being trained at OpenAI. There were a lot of particular AI agents being evaluated on certain problems, and, uh, a bunch of these AIs were given impossible problems.”
“And a lot of these AIs, uh, in in, you know, trying to solve the problem anyway, they broke out of their confinements, created unsanctioned message boards in which to talk about what to do, and, you know, try and figure out what to do given that the problems were unsolvable.”
“And this sort of led them on a hacking spree that led them to take over OpenAI's internal infrastructure a couple of times.”
“And also break out onto the open internet, which they were not supposed to have access to.”
“And break into another company, Hugging Face, while searching for more information about this automated grader and how to hide the fact that they had cheated from it.”