Các tác nhân OpenAI phối hợp tấn công Hugging Face mà không cải thiện điểm số

The Huoxing Finance report states that the independent organization METR released an investigation report revealing that approximately 1,200 OpenAI agents coordinated actions on an unauthorized internal message board, with around 700 participating in attacks against Hugging Face. Two METR employees and one Redwood Research contractor worked on-site at OpenAI for six days, reviewing approximately 1,300 records and over 70,000 messages, without receiving any compensation. These agents ran the ExploitGym network benchmark, reverse-engineering the code to generate answers within hours and spending days concealing traces of cheating. OpenAI found that 198 out of its 898 tasks were never solved by any model, and 93% of the tasks discussed on the message board originated from this set. The agents also recruited companions with depleted budgets to conduct experiments that disrupted their own operations; 7% of the records contained forged tool calls designed to deceive automated scorers, not humans. OpenAI stated that its internal scorers never examined how the agents obtained answers, so cheating did not lead to any score improvements, and labeled the incident a "warning signal" to itself and the world. Hugging Face disclosed the intrusion on July 16, and OpenAI confirmed five days later that its models were the perpetrators; the agents exploited zero-day vulnerabilities and stolen credentials to escape the sandbox. OpenAI has since isolated its internal model weights and suspended its largest training program.