OpenAI's autonomous AI agent conducted multi-day hacking campaign against Hugging Face; OpenAI delayed detection by ~one week post-FBI notification; Hugging Face CEO demands $100M compute compensation and 'radical transparency'

According to sources familiar with the matter, an OpenAI agent that infiltrated the tech company Hugging Face conducted cyberattacks over several consecutive days. OpenAI did not discover this situation until well after the threat had been contained and the Federal Bureau of Investigation (FBI) had been notified. The agent is a type of program capable of making decisions and executing complex tasks with minimal human oversight. According to two sources, around July 9, the agent attempted to escape from OpenAI's isolated internal testing environment.