Alibaba Bans Claude for Employees Amid China Spying Code Concerns

Alibaba has reportedly banned its employees from using Anthropic’s Claude Code at work and instructed them to switch to the Chinese company’s own Qoder coding platform.

The decision follows scrutiny of Claude Code features that can inspect user environments and potentially identify people connected to China. It also marks a further escalation in the dispute between Alibaba and Anthropic over the alleged use of Claude to improve competing Chinese AI models.

Claude Code Blocked

Claude Code is Anthropic’s AI coding assistant for software developers. Despite Anthropic restricting access for users and companies in China, the service has become popular among Chinese programmers.

A person familiar with Alibaba’s internal order told Reuters that employees were being directed to replace Claude Code with Qoder, Alibaba’s agentic coding platform. Qoder offers tools for code completion, autonomous development tasks and understanding large software projects.

Neither Alibaba nor Anthropic immediately commented on the reported ban. Alibaba has also not publicly responded to Anthropic’s earlier accusations against the company.

China-Tracking Concerns

The ban emerged shortly after developers reported that Claude Code contained mechanisms capable of examining information from a user’s environment.

The tool reportedly checked details including timezone and proxy-related information. Developers also said it could insert subtle markers into prompts before sending them to Anthropic’s servers. These mechanisms could help Anthropic identify users attempting to access Claude from restricted locations or through intermediaries.

An Anthropic employee said on X that the feature was an experiment introduced in March. According to the employee, it was designed to combat account abuse by unauthorized resellers and protect Claude against model-distillation attempts.

Anthropic Accuses Alibaba

The workplace ban comes amid a growing dispute between the two companies.

Anthropic recently accused Alibaba of conducting a large-scale “distillation” effort against Claude. Model distillation involves training a smaller or less capable AI model using answers generated by a more advanced system.

Anthropic claimed Alibaba had illicitly used Claude’s outputs to obtain its capabilities more quickly. It said the effort could accelerate the development of Chinese models toward the performance of its advanced Mythos Preview system. The accusation appeared in a letter sent to two US senators and reviewed by Reuters.

Alibaba has not publicly responded to the allegation.

Restrictions Are Difficult to Enforce

Anthropic does not officially allow companies and users in China to access its services. Its restrictions also cover overseas entities controlled by companies based in unsupported regions.

The company has said such businesses could use Claude to improve their own AI systems through techniques such as distillation, even when accessing the service through subsidiaries based in other countries.

However, restrictions on individual users can be difficult to enforce. A person familiar with Alibaba’s ban said users can operate servers in the United States and make their traffic appear to originate there.

Businesses face greater legal, security and compliance risks, making them more likely to stop using restricted foreign AI services.

Alibaba Pushes Qoder

Alibaba employees are now being directed toward Qoder, the company’s own AI development platform.

Qoder includes an AI-powered development environment, autonomous coding agents and tools that can examine and work across an entire codebase. Alibaba also offers a China-based Qoder suite designed to meet local data security and compliance requirements.

The switch gives Alibaba greater control over employee data and reduces its reliance on an American AI provider that openly restricts Chinese access.