China Warns of 'Security Backdoor' in Anthropic's Claude Code; Calls for Uninstall

Summary

China's National Vulnerability Database (NVDB), affiliated with the Ministry of Industry and Information Technology, issued a warning Wednesday of a 'security backdoor' embedded in Anthropic's Claude Code that could transmit sensitive user information—including locations and identity-related identifiers—back to Anthropic's servers without consent. The NVDB advised users and institutions to conduct immediate security checks and uninstall the tool or upgrade to the latest secure version.

Anthropicresponded by characterizing the disputed mechanism as a standard anti-abuse measure designed to detect fraud and prevent unauthorized access from restricted regions. Claude Code is an AI agent capable of generating, debugging, and reviewing code based on user prompts. Although Anthropic prohibits users in China and other adversarial nations from accessing the tool, some circumvent these restrictions using VPNs or proxy services.

The controversy underscores escalating geopolitical tensions in AI. Chinese tech giant Alibaba announced last week it would ban Claude Code usage among employees effective July 10 due to security concerns—part of a broader pattern of friction between Anthropic and China's tech sector. Anthropic has previously accused Alibaba of reverse-engineering its models through distillation. Claude Code engineer Thariq Shihipar acknowledged the detection mechanism as an anti-distillation experiment launched in March, stating the team had developed stronger mitigations and planned to fully roll back the feature.