China flags alleged Claude Code backdoor risks

The Wall Street Journal reports that China's National Vulnerability DataBase warned of alleged backdoor vulnerabilities in versions of Anthropic's Claude Code. The warning said the tool could transmit sensitive data without user consent and advised users to uninstall or update. Anthropic said the disputed mechanism was part of an anti-abuse experiment tied to distillation and unauthorized reselling, not a backdoor.

Key details: China's vulnerability database warned about Claude Code, The warning claimed sensitive data could be transmitted without consent, Anthropic said the mechanism was an anti-abuse experiment.

Why it matters: AI coding tools are now part of the U.S.-China security fight, with anti-abuse telemetry itself becoming politically contested.