微软研究院发布研究,指出即使单个AI代理是安全的,互联代理的生态系统也不一定安全,需要新的方法来应对网络级风险。
Safe agents don’t guarantee a safe ecosystem of interconnected agents. Microsoft Research examines what breaks when AI agents interact and why network-level risks require new approaches. Learn more: https://t.co/FngPJsamPT https://t.co/X40wF9IH1R
likes: 59 | retweets: 13 | replies: 3 | views: 9820