The Chinese Ministry of National Security has released a safety manual for lobster farming According to BlockBeats, on March 17th, the Chinese Ministry of National Security issued usage tips for the recently popular open-source AI agent tool OpenClaw (nicknamed "Lobster"), stating that while this type of highly authorized AI agent may bring efficiency improvements, it may also pose security risks such as host control, data leakage, and information abuse. It is recommended that users use it rationally and strengthen security protection. The prompt points out that "Lobster" can directly execute user instructions to complete tasks by integrating communication software and large language models, and combining them with high permission system access capabilities, thus achieving a transition from "providing suggestions" to "automatic execution". At the same time, its plugin system supports multiple functions such as file management, email writing, schedule scheduling, and web browsing, and has long-term memory and the ability to actively trigger tasks, so it is called "raising lobsters" by some users. Security agencies remind that if not configured properly, intelligent agents may pose multiple risks, including: remote takeover of devices due to high privilege operation, theft of sensitive data, use of social accounts to publish false information, and attacks on systems through malicious plugins. To reduce risks, it is recommended that users follow the principle of minimum privilege when running intelligent agents, encrypt sensitive data, and deploy them in isolated environments such as virtual machines or sandboxes. At the same time, regularly check the source of plugins, access permissions, and system logs to avoid exposing core devices directly to the public network environment.