China's National Security Authority Issues Safety Guidance on OpenClaw AI Agent Software
China's National Security Ministry released a safety guide warning of four major security risks with OpenClaw, a widely-used AI agent software. The risks include system compromise, data theft, statement manipulation, and technical vulnerabilities, particularly from malicious plugins that could steal sensitive device information.
What happened
China's National Security Ministry released a "safety guide for raising lobsters" on OpenClaw, a popular AI agent software known by the nickname "ιΎθ¦" (lobster). The guide identified four major security risks: hosts potentially being compromised, data being stolen, statements being altered, and technical vulnerabilities.
Reported impact
The guide warned that installation of malicious plugins from untrusted sources could lead to theft of sensitive information from local devices. If personal sensitive data is entrusted to OpenClaw and the system is breached, privacy could be exposed.
Index analysis
OpenClaw operates by integrating communication software with large language models and executing user commands with high permissions remotely through chat applications. The system includes many built-in plugin skills that users can download, creating a tool platform covering file management, email writing, calendar scheduling, web browsing, and scheduled tasks.
Governance controls
The National Security Ministry recommended that users check whether control interfaces are exposed to the public network, whether permission settings are too high, whether stored credentials have been leaked, and whether plugin sources are trustworthy. Users should encrypt sensitive data and strictly limit its operational scope.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.