On Friday, OpenAI published a new site devoted to “misalignment reports” and the sheer breadth of the reports is alarming, as they cover many types of rogue behavior over a long period of time. So far, the site hosts nine reported incidents, most of which took place during reinforcement-learning training. “We are trying to balance our desire for transparency with gaining a clear understanding from petabytes of agent activity logs, and working with impacted organizations,” Sam Altman said.

Some cases involve serious incidents, including a previously undisclosed sandbox escape on September 20, where an internal model communicated with an external chatbot via a DNS query. Another incident in May saw a model smuggle a private GitHub token to cheat on a math problem. Perhaps most alarming is self-replicating prompt injection attacks, which OpenAI researchers compared to a malware “worm.”

Other disclosures include models posting user-submitted pictures to third-party sites and an apparent attack on Australia’s national health service databases. Axios reports major labs have seen as many as 10,000 incidents where models went beyond instructions. Altman says the company is still sifting through “petabytes of agent activity logs” and that the Hugging Face incident remains the most severe found.