China's National Vulnerability Database (NVDB) has issued a warning regarding potential security vulnerabilities in Anthropic's Claude Code AI models, according to reports from Reuters. The database, which is associated with the Chinese government, claims that recent versions of Claude Code contain a monitoring mechanism that could transmit sensitive user information to remote servers without explicit consent.
The reported issue specifically affects versions of Claude Code from 2.1.91 up to 2.1.196. Version 2.1.196 was superseded by version 2.1.200 on July 3. The NVDB is recommending that users either uninstall the affected versions or update to the latest available release to mitigate any potential risks. The agency also urged users to enhance their data control and monitoring practices for information accessed by Claude.
This development follows a separate incident last week where Alibaba reportedly banned its employees from using Claude for work purposes, citing concerns over potential user identification. The situation between Anthropic and Chinese entities has become increasingly strained, with Anthropic previously accusing Alibaba of harvesting data by using Claude through fraudulent accounts. Anthropic has also been reportedly taking steps to prevent Chinese companies from circumventing its access restrictions through overseas subsidiaries or VPNs.
The NVDB's warning suggests a growing trend of Chinese agencies favoring domestically developed AI models. Anthropic has not yet responded to requests for comment on this story.