Alibaba has banned its employees from using Anthropic‘s AI coding tool, Claude Code, following the discovery of hidden tracking software used to identify users in China.

The internal restriction, reported by The Next Web, takes effect on 10 July.

The decision follows growing tensions between the two technology companies over intellectual property.

Alibaba added Claude Code to its list of high-risk software due to security vulnerabilities, recommending that employees switch to Qoder, its proprietary coding platform.

A developer discovered the tracking mechanism in late June after analyzing the software. The hidden markers had been active since April, checking system timezones and scanning proxy URLs against Chinese domains and AI laboratories.

Data compliance and developer access

The discovery has raised significant transparency and cross-border data compliance concerns within the developer community.

Chinese cybersecurity firm Huorong Security noted that the tracking mechanism introduces operational and legal risks for companies.

This development is particularly relevant for financial firms and institutions operating across Hong Kong and the mainland.

Regulators in these jurisdictions strictly enforce data sovereignty rules, making hidden data tracking a critical compliance risk.

Because Claude Code requires deep access to a local file system to modify and execute code, developers expressed concern over how user data was handled.

The mechanism used hidden signals within system prompts sent back to Anthropic’s servers.

Disputes over model distillation

Anthropic engineer Thariq Shihipar clarified on X that the tracking was part of an experiment launched in March.

He stated the system was designed to prevent account abuse from unauthorized resellers and protect the company’s models against distillation.

Distillation involves using the outputs of a powerful AI model to train a smaller, less expensive one.

Anthropic has actively campaigned against this practice, framing it as a threat to the business models of frontier AI organizations.

In June, Anthropic informed the US Senate Banking Committee that entities linked to Alibaba’s Qwen AI lab had conducted a massive distillation campaign against Claude using thousands of fraudulent accounts. Alibaba has denied these allegations.

Shifting to domestic tools

The controversy accelerates a broader push by Chinese technology firms to reduce their reliance on American AI software. Corporate entities increasingly view foreign tools as carrying unpredictable regulatory and access risks.

Lizzi Lee, a fellow at the Asia Society Policy Institute’s Centre for China Analysis, noted that the conflict highlights how the technology competition has moved into access control and sovereignty.

“If a US AI coding tool can detect Chinese usage or proxy access, then it is not surprising for major Chinese tech companies to not want employees using it internally,”

said Lee.

The restriction gives Alibaba further justification to integrate its own Qwen models across its business units.

By restricting external tools, the firm is strengthening its domestic AI ecosystem.

Featured image credit: Edited by Fintech News Hong Kong, based on image by pvproductions via Magnific