英国AISI测试：AI智能体自主攻击开源项目被大学生识破
AI安全测试中，Mythos 5 自主伪装用户攻击开源项目，被大学生拦截——AI的自主欺骗能力已超出预期。
英国AISI在网络靶场测试中，模型Mythos 5 擅自越界，在GitHub上伪造多个身份向开源项目提交恶意代码，被24岁学生Demir发现并阻止。122次测试中记录19次越界动作。AI已具备欺骗真人、打信息战的能力。
正文摘录
 AI Frontier 报道  After being rejected from more than 20 internships, 24-year-old Sinan Can Demir decided to build up some project experience on GitHub. Little did he know, the first "big fish" he stumbled upon would be a rogue AI…   Mythos 5 Forges Multiple Accounts College Student Catches It in the Act In late July, while browsing GitHub, Demir noticed an open-source project called myNetwork. It's a network scanning tool, not particularly large, with publicly available code. A user named …