China's Ministry of State Security releases manual on safe lobster farming.

According to Mars Finance, on March 17th, China's Ministry of State Security issued a usage warning regarding the recently popular open-source AI agent tool OpenClaw (nicknamed "Lobster"). The warning stated that while such high-privilege AI agents can improve efficiency, they may also pose security risks such as host control, data leaks, and information misuse. Users are advised to use it rationally and strengthen security precautions. The warning pointed out that "Lobster," by integrating communication software and a large language model, along with high-privilege system access capabilities, can directly execute user commands to complete tasks, thus transforming from "providing suggestions" to "automatic execution." Furthermore, its plugin system supports various functions such as file management, email composition, scheduling, and web browsing, and possesses long-term memory and the ability to proactively trigger tasks, leading some users to refer to it as "raising lobsters." Security agencies warn that improper configuration can lead to multiple risks, including: high-privilege operation resulting in remote device takeover, the theft of sensitive data, the misuse of social media accounts to spread false information, and attacks on systems through malicious plugins. To mitigate risks, users are advised to run intelligent agents according to the principle of least privilege, encrypt sensitive data, and deploy them in isolated environments such as virtual machines or sandboxes. At the same time, users should regularly check the source of plugins, access permissions, and system logs to avoid exposing core devices directly to the public network environment.