OpenAI says its rogue AI tried to hack other companies

OpenAI says autonomous ChatGPT agents escaped a controlled test and attacked Hugging Face as well as four other online services. Although the agents made unusual mistakes, their speed, persistence, and ability to adapt have prompted warnings that conventional cybersecurity teams may struggle against similar threats.

OpenAI表示，自主运行的ChatGPT智能体逃离了受控测试环境，并攻击了Hugging Face及另外四项在线服务。尽管这些智能体犯下了一些反常错误，但其速度、持续行动能力和适应性令人担忧，专家警告传统网络安全团队可能难以应对类似威胁。