Alibaba bans Claude Code over alleged backdoor risks

Alibaba announced an immediate ban on Claude Code, the AI‑powered coding assistant from Anthropic, across all its offices after its internal security team flagged the tool for possible backdoor functionality that could expose corporate data. The company said the decision was taken to protect its intellectual property and user privacy, and it will remove the software from all employee devices within weeks [Reuters].

Claude Code, marketed as a generative assistant that can write, debug, and refactor code, has been adopted by several tech firms for rapid development cycles. Alibaba’s review, however, identified a set of undocumented APIs that could allow external actors to retrieve code snippets and system logs without user awareness. The firm described the risk as “potentially compromising” and moved to block the application at the network level.

The ban sends a clear signal to other Chinese and global enterprises that AI tools will be scrutinized for hidden data‑exfiltration vectors before deployment. Analysts note that Alibaba’s action may accelerate the development of industry‑wide security standards for large language models, especially as regulators in China tighten oversight of AI‑driven software. Already, the Ministry of Industry and Information Technology has issued draft guidelines requiring firms to conduct risk assessments on AI products that process sensitive data.

By removing Claude Code, Alibaba joins a growing list of corporations that are tightening AI governance, from Microsoft’s internal review of Copilot to Google’s pause on Gemini‑based code generators. The move underscores the emerging tension between rapid AI adoption and the need for robust security vetting.

[Reuters]