China's National Vulnerability Database has issued a severe security warning, claiming the US AI company Anthropic's coding tool, Claude Code, contains a "security backdoor."
The regulator alleges the tool could transmit sensitive user information, including location and identity data, to Anthropic servers without consent.
Chinese tech giant Alibaba has informed employees it will ban internal use of Claude Code starting July 10th due to these security concerns.
Anthropic blocks access from China, but the tool is still used there via VPNs. The company has not commented on the specific allegations.
An Anthropic engineer stated on social media that a data collection feature was an experiment to combat abuse and account reselling, and that it was being phased out.