China’s CERT warns OpenClaw can inflict nasty wounds

China’s national CERT warned that the agentic AI tool OpenClaw has weak default security, multiple disclosed severe vulnerabilities (including ones that could enable credential theft), and is susceptible to attacks via malicious web instructions and poisoned plugins. The advisory recommends isolating OpenClaw in containers or nonproduction VMs, restricting management ports and plugin access, enforcing strict authentication, and disabling automatic updates; the tool’s rapid uptake and one-click cloud deployments prompted some government bans.