Safety and Security: China vs Anthropic’s Claude Code
Alert: China’s National Vulnerability Database (NVDB), a cybersecurity platform tied to the industry ministry, issued an alert claiming certain versions of Anthropic’s coding tool Claude Code contain a “security backdoor.”
Alleged issue: According to NVDB, versions between 2.1.91 (April 2) and 2.1.196 (June 29) include a built-in monitoring mechanism that can transmit sensitive information—like user location and identity-related identifiers—to remote servers without consent.
Guidance: NVDB advised institutions and developers to immediately audit deployments and either uninstall affected versions or upgrade to newer releases where the monitoring code is reportedly removed.
Geopolitical angle: The alert also plays into US-China tensions over AI, as China highlights security risks in US-developed tools while asserting its own cybersecurity oversight role.
