China Warns of Security Risks in Anthropic's AI Coding Tool

Kathmandu. A Chinese cybersecurity agency has warned users to be cautious about security risks in some versions of the AI coding tool 'Cloud Code' by American artificial intelligence (AI) company Anthropic.

The National Vulnerability Database (NVDB), affiliated with China's Ministry of Industry and Information Technology, has stated that a 'security backdoor' in the tool can send sensitive user location and identification information to the company's servers without user consent.

'Cloud Code' is an AI coding tool that has the ability to generate computer code based on user instructions, fix software errors, and review code.

NVDB has urged the relevant organizations and users to immediately conduct security testing of the tool's usage and, if necessary, uninstall it or update it to the latest version with security risks removed. It also suggested strengthening network traffic monitoring to prevent sensitive information leakage.

According to AFP, Anthropic did not immediately respond to requests for comment on these allegations. Meanwhile, Chinese tech company Alibaba has announced its decision to ban the use of 'Cloud Code' from July 10 due to security concerns.

Meanwhile, Thariq Shihipar, a 'Cloud Code' engineer at Anthropic, said via social media platform X that the data tracking feature in the system was part of a test to prevent unauthorized resellers and misuse of AI models. He informed that the process of removing the feature has already begun and it will be completely removed in the new version.