A security agency under the Chinese government said it found security vulnerabilities in Anthropic's artificial intelligence (AI) coding tool "Claude Code" and recommended deletion, among other steps. But Anthropic said Chinese corporations are not eligible to access Claude Code.
CNVDB, a cybersecurity vulnerability databases run by the Chinese government, warned on the 8th that "multiple versions of Claude Code released from April to June can transmit sensitive information, such as a user's location and identity, to a remote server without user consent due to built-in monitoring mechanisms."
It added that "such mechanisms can pose a serious threat," advising users to delete the software or update to the latest version with the related code removed. In connection with this, China's big tech Alibaba notified employees it will ban the use of "Claude Code" during work starting on the 10th.
The Wall Street Journal (WSJ) analyzed that China's move heightened tensions in its competition with the United States over AI supremacy. Anthropic did not immediately respond to the Chinese agency's statement. But Anthropic had previously said that Chinese corporations such as Alibaba are not eligible to access Claude services.
China's latest action appears related to a post on the global online community site Reddit last week that claimed Anthropic secretly planted code in its software to identify users accessing from China.
Regarding the claim, an Anthropic employee said on the social media platform X (formerly Twitter) that the code was part of an experiment that began in March, explaining it was "to prevent account abuse by unauthorized resellers and protect against so-called 'distillation (model cloning and training).'"
China has not approved Anthropic's services, and Anthropic has restricted access to Claude in China for national security reasons. But Chinese researchers and engineers are reportedly using Claude through various workarounds.