Anthropic's AI coding assistant, Claude Code, faces new regulatory scrutiny after Chinese authorities alleged that several versions of the software contain a ‘backdoor’. Authorities claim that the flaw could allow sensitive user information to be transmitted to remote servers. The warning also highlights the growing geopolitical contest over AI software security and enterprise adoption.
On July 8, China's National Vulnerability Database (NVDB), overseen by the Ministry of Industry and Information Technology, issued the alert. The advisory claims affected versions of Claude Code could send identity-related information and user location data to Anthropic without explicit consent. Regulators urged organizations to update or discontinue the software immediately.
Anthropic has strongly disputed the allegation. It claimed that the disputed code was for an experiment aimed at checking for any abuse of the system and detecting unauthorized access, model distillation, and resale of Claude services. It argued that it was not meant to be used as a backdoor.
However, Claude Code has not been made available officially in China, and the mechanism has been removed from the software. The controversy has already affected enterprise adoption. Alibaba has instructed employees to stop using Claude Code as of July 10, citing security concerns about the tool. Staff have reportedly been directed to use Alibaba's in-house AI coding assistant instead.
The episode underscores how AI security has become as important as model performance. Governments are becoming more focused on how AI tools collect telemetry data, authenticate users, and transmit data across national boundaries.
From the enterprise perspective, the dispute becomes an illustration of how decision-making concerning the implementation of AI has moved beyond efficiency. Issues of regulatory compliance, software transparency, and national security have become crucial considerations in vendor selection.
Whether China's allegations are technically justified or not, the dispute highlights the growing importance of cybersecurity, geopolitics, and AI. As governments increase scrutiny of foreign AI software, vendors could face growing pressure to explain how their products handle telemetry systems and user data.
Read More: Anthropic Unveils Claude Cowork AI Tool for Smarter Workplace Automation