Alibaba Bans Claude Code Over Security Concerns

Alibaba Bans Claude Code Over Security Concerns

Quick Reads

• Alibaba will ban Claude Code from workplace use starting July 10.
• The company cited alleged security and backdoor risks.
• Employees are being directed to use Alibaba’s Qoder coding assistant.
• The move adds to growing US-China tensions over AI development.

When developers choose an AI coding assistant, they expect speed, accuracy, and trust. This week, trust became the bigger story. Alibaba has reportedly ordered employees to stop using Anthropic’s Claude Code at work, marking another flashpoint in the increasingly tense AI rivalry between China and the United States.

The ban will take effect on July 10 after Alibaba classified Claude Code as high-risk software over alleged backdoor security concerns. The company has instructed employees to switch to its in-house coding assistant, Qoder, instead. Neither Alibaba nor Anthropic immediately responded to requests for comment.

The decision follows days of online scrutiny after developers claimed Claude Code contained hidden mechanisms capable of identifying users linked to China. The feature was introduced as an experiment to prevent account abuse by unauthorised resellers and reduce AI model distillation. The company also said stronger protections have since replaced the experiment.

The move arrives just weeks after Anthropic accused Alibaba of extracting Claude’s capabilities through large-scale AI model distillation, an allegation that further strained relations between the two companies. While those claims remain disputed, the latest restriction highlights how security, intellectual property, and geopolitics are becoming inseparable from AI software development.

The story goes beyond one coding tool. Enterprise customers are paying closer attention to how AI applications collect data, verify users, and protect sensitive information. As governments and technology companies tighten security policies, developers may increasingly find themselves choosing tools based not only on performance, but also on trust.