PANews reported on March 17 that the open-source AI agent tool OpenClaw ("Lobster"), while integrating communication software and large models to autonomously execute tasks with high privileges, poses risks such as host takeover, data leakage, tampering of speech, and technical vulnerabilities. The article points out that users often grant "Lobster" the highest system privileges and entrust it with handling sensitive data. If attackers use malicious plugins to poison or remotely control it, the device could be illegally manipulated, resulting in privacy and financial losses, or even being used to spread false information and commit fraud. The Ministry of State Security recommends that users conduct a security check on "Lobster," adhere to the principle of least privilege, run it in an isolated environment, encrypt sensitive data, and treat it as a "digital employee" requiring compliant, secure, and controllable management.