China’s cybersecurity authorities warn that certain versions of Anthropic’s Claude Code contain security issues that may expose sensitive user information. Multiple outlets report that China says the affected releases can transmit data such as user location and user identity to remote servers without the user’s consent. The concern centers on a built-in monitoring mechanism described as functioning like a backdoor. The reports cite China’s National Vulnerability DataBase and specify that the alert covers Claude Code versions 2.1.91 through 2.1.196. The guidance presented by sources urges users to uninstall the application or upgrade to a safer version. The warnings arrive against a backdrop of heightened US-China scrutiny of AI-related technologies. One outlet also notes that Alibaba has restricted employees from using Claude Code at work, reflecting broader tensions around use and governance of AI tools. Anthropic and any detailed technical assessment beyond China’s disclosure are not included in the provided summaries.