China Flags Security Concerns with Anthropic's AI Coding Tool Claude Code

China's National Vulnerability Database (NVDB) has issued a warning about security backdoor vulnerabilities in Anthropic's popular AI coding tool Claude Code. The issue affects versions of the tool released between April and June 2026, which can send sensitive information such as user location and identity to remote servers without the user's consent.

According to the NVDB, the problem stems from a built-in monitoring mechanism that was intended to prevent account abuse from unauthorized resellers and protect against distillation. However, Anthropic engineer Thariq Shihipar has confirmed that this was an experiment launched in June, which has since been deemed a security risk.

Anthropic has already restricted the use of its AI tools in China due to national security risks. The company has also accused Chinese AI labs of distilling Claude Code, creating 24,000 fraudulent accounts to train smaller models. In response to the Chinese government's warning, Anthropic has promised to roll back the problematic feature in its next release.

The controversy surrounding Claude Code highlights the ongoing debate about the security and ethics of AI development. As AI tools become increasingly sophisticated, it is essential to ensure that they are designed with robust security measures in place to prevent potential risks and vulnerabilities.