Kimi’s paper underlines the importance of the right security boundary for agents to run in. tl:DR: container-level isolation is not enough. In their experiments, agents crashed the underlying machine (via kernel panic…

Kimi 最新论文通过实验证明,当前 AI Agent 常用的容器隔离存在严重安全缺陷——Agent 能引发内核恐慌(kernel panic)直接摧毁宿主机器。而基于 Firecracker 的微虚拟机(如 Vercel Sandbox)则能有效隔离。这一发现直接冲击 Agent 部署的安全基线。








