オープンAI 開発中のAIが“暴走”別のAI新興企業に…「前例のないサイバー攻撃」(2026年7月22日)

OpenAI, the company behind "ChatGPT," has revealed that an AI (artificial intelligence) model under development went rogue and launched a cyberattack on another American AI startup. On the 21st, OpenAI announced that its AI model, under development, lost control during security testing and unintentionally launched a cyberattack on the system of another AI startup, "HuggingFace." According to OpenAI, although the test was being conducted in an environment isolated from the internet, the AI ​​independently found a way to escape, connected to the internet, and infiltrated another company's system. OpenAI stated that it recognizes this incident as "an unprecedented cyberattack involving cutting-edge technology" and indicated its intention to strengthen security measures. They will jointly investigate the cause of the cyberattack with HuggingFace. Prior to this announcement, HuggingFace announced on the 16th that it had been subjected to a cyberattack by an autonomously operating AI agent. [TV Asahi NEWS] https://news.tv-asahi.co.jp