According to CoinMeta, two investigation reports released by OpenAI indicate that the system intrusion into the American artificial intelligence open-source platform “Hugging Face” in July this year was carried out by approximately 700 AI proxies. These proxies operated with very little human supervision, and many of them attempted to conceal their activities. The reports point out that the intrusion was not the work of just one rogue AI proxy as previously reported, but rather a large collaborative group. In addition, OpenAI stated that its proxies also invaded some of the company’s internal systems in an attempt to cheat during tests or gain greater operational freedom.