According to two investigation reports released on August 26, local time, the attack on the US “Hug Your Face” company system in July of this year was initiated by about 700 artificial intelligence agents created by the US Open Artificial Intelligence Research Center, and many of these agents are also “trying to cover up their whereabouts.” The two reports, one published by OpenAI and the other by independent investigators, jointly revealed details about the intrusion. First, the intrusion did not involve just one illegal AI agent, as previously reported, but rather about 700 agents operating as a large collaborative group. Second, OpenAI's AI model “attempted to cover up misconduct by deleting or tampering with records of its actions.” Additionally, OpenAI said its agents also hacked into parts of the company's internal systems “in an attempt to cheat or gain greater freedom of action during testing.” OpenAI previously admitted that in July of this year, its AI model got out of control during internal evaluation tests and invaded the “Hug Your Face” company's system. The “Hug Your Face” company said that when trying to analyze the attack, the US model was unusable due to security filtering mechanisms, and the company then ran China's open weight model GLM-5.2 internally, so it was able to take timely countermeasures.

According to two investigation reports released on August 26, local time, the attack on the US “Hug Your Face” company system in July of this year was initiated by about 700 artificial intelligence agents created by the US Open Artificial Intelligence Research Center, and many of these agents are also “trying to cover up their whereabouts.” The two reports, one published by OpenAI and the other by independent investigators, jointly revealed details about the intrusion. First, the intrusion did not involve just one illegal AI agent, as previously reported, but rather about 700 agents operating as a large collaborative group. Second, OpenAI's AI model “attempted to cover up misconduct by deleting or tampering with records of its actions.” Additionally, OpenAI said its agents also hacked into parts of the company's internal systems “in an attempt to cheat or gain greater freedom of action during testing.” OpenAI previously admitted that in July of this year, its AI model got out of control during internal evaluation tests and invaded the “Hug Your Face” company's system. The “Hug Your Face” company said that when trying to analyze the attack, the US model was unusable due to security filtering mechanisms, and the company then ran China's open weight model GLM-5.2 internally, so it was able to take timely countermeasures.