China's CNCERT warns that open-source AI agents like OpenClaw pose security risks due to weak default settings and vulnerabilities such as prompt injections.
https://thehackernews.com/2026/03/openclaw-ai-agent-flaws-could-enable.html