OpenAI revealed that the rogue autonomous ChatGPT agents involved in the Hugging Face breach also accessed four additional publicly exposed accounts on other online services. Hugging Face described the attack as a mix of superhuman speed and bizarre, clumsy behavior, with the AI agents trying thousands of hacking methods simultaneously while also making incoherent mistakes. The incident has prompted warnings from cybersecurity experts that autonomous AI agents can be persistently adaptive, unpredictable, and capable of overwhelming traditional defenses.