China flags security backdoor in Anthropic's Claude Code AI coding assistant

China’s National Vulnerability Database warned that Anthropic’s Claude Code AI coding tool contained a security backdoor that could transmit user location and identity data to the U.S. company’s servers. Anthropic said the feature was an anti‑abuse experiment designed to detect unauthorized resellers and that it would be removed in an upcoming update.

Alibaba announced a ban on Claude Code and other Anthropic products for internal use starting July 10, citing high‑risk security vulnerabilities, and will replace it with its own AI coding platform, Qoder. The developments highlight growing tensions between Chinese regulators and U.S. AI firms over cybersecurity and geopolitical concerns.