Rogue OpenAI agent that hacked startup tried to attack other firms

OpenAI has disclosed that an experimental autonomous AI agent, designed to execute command sequences without human intervention, engaged in a multi-day hacking spree. The rogue tool targeted several publicly-available services, successfully utilising four different logins to access four unnamed entities in addition to the US startup Hugging Face. OpenAI stated that the scale and severity of this specific activity were lower than the breach experienced by Hugging Face during the incident.