Alibaba Blacklists Anthropic's Claude Code Over Hidden Tracking and Security Concerns

Alibaba Blacklists Claude Code Over Backdoor Allegations

Alibaba Group Holding has officially banned its employees from using Anthropic's Claude Code for work, classifying the AI coding assistant as high-risk software. According to an internal notice seen by the South China Morning Post, the Chinese technology giant will prohibit all staff members from using the tool in office environments starting on July 10, 2026. The restriction was first reported by the Chinese financial news outlet Yicai before being confirmed by international sources. In place of the banned software, Alibaba is directing its developers to use its own proprietary platform, Qoder.

The decision to blacklist the tool follows an internal evaluation triggered by security researchers on platforms like Reddit and GitHub, who discovered hidden tracking code within the software. The internal notice stated that Claude Code was found to carry backdoor risks and security vulnerabilities. Specifically, researchers revealed that the software covertly inspected user environments to determine if they were connected to China.

The Mechanics of Anthropic's Covert Tracking

The controversy centers on how Claude Code identifies and flags users. Security findings show that the software was designed to quietly check whether a user's system timezone matched Chinese time zones. Additionally, it scrutinized proxy URLs to see if they matched specific Chinese domains or known domestic artificial intelligence laboratories. Instead of displaying a standard warning or compliance notice to the user, the software allegedly injected subtle, machine-readable markers into the prompts sent back to Anthropic's servers, effectively marking certain users without their knowledge.

An Anthropic employee addressed the findings on the social media platform X, confirming that the tracking mechanism was an experiment launched in March. According to the company, the feature was designed to prevent account abuse, stop unauthorized reselling of its services, and defend against unauthorized model distillation.

Data Scraping Accusations and the Battle Over Distillation

This workplace ban is the latest escalation in a bitter, ongoing dispute between the two technology companies. Just two weeks prior to the ban, Anthropic sent a letter to two U.S. senators accusing Alibaba of executing a massive model distillation campaign. In the letter, Anthropic alleged that Alibaba used 25,000 fake accounts to extract 28.8 million interactions from the Claude model. This distillation process involves training a less capable model on the outputs of a superior one, which Anthropic claims was done to accelerate China's progress toward achieving Anthropic's advanced Mythos Preview capabilities.

The conflict highlights the broader geopolitical struggle to dominate the artificial intelligence sector. Anthropic has been actively working to seal loopholes that allow Chinese entities to access Claude, despite official bans. While Chinese developers frequently bypass geographic restrictions using virtual private networks, cloud platforms, and proxy transfer stations, industry observers note that large corporations face much higher legal and compliance risks, making them far more sensitive to these covert tracking mechanisms.

This escalating cycle of covert tracking and corporate blacklisting suggests that the technical friction between American AI safeguards and Chinese development ambitions will inevitably drive a complete decoupling of their respective software ecosystems.