Chinese tech giant Alibaba has imposed a categorical ban on employees using Anthropic's Claude Code tool. This decision comes as a direct response to recent accusations by the American AI developer against the Chinese corporation, as well as the discovery of mechanisms in Claude Code capable of identifying users associated with China.
Covert Surveillance or Intellectual Property Protection?
Claude Code is an advanced AI assistant for programmers that, despite Anthropic's official restrictions on access from China, has gained significant popularity among developers in the country. It is this popularity that appears to have led to the crisis.
Developers discovered that Claude Code checks the user's environment, including time zone and proxy data, and then inserts hidden markers into requests sent to Anthropic's servers. Anthropic itself explained that this was an "experiment launched in March" to protect against unauthorized resellers and prevent model distillation — a process where a less powerful model is trained on the responses of a more powerful one.
However, Alibaba interpreted this differently. The ban on using Claude Code came literally a few days after Anthropic sent a letter to two U.S. senators, accusing Alibaba of attempting to unauthorizedly extract the capabilities of its model. Anthropic claims that Alibaba used "distillation" methods to accelerate China's access to the advanced capabilities of the Mythos Preview model.
Escalation of Technological Tensions
This incident is just part of a broader conflict. Alibaba's management has instructed employees to use its own Qoder programming platform, a logical step amid escalating confrontation.
Anthropic, for its part, is tightening restrictions by closing "loopholes" (VPNs, cloud platforms, and proxies) that allowed Chinese companies to access Claude. Restrictions targeting China were difficult to enforce against individual users who could deploy servers in the U.S. and mask traffic. However, for corporations, the legal and compliance risks are now becoming too high.
Against the backdrop of American developers' attempts to prevent unauthorized access, Chinese cloud and AI companies are actively transitioning to domestic and open-source models — such as DeepSeek, Alibaba Qwen, Moonshot, and Zhipu. Notably, these same Chinese models are simultaneously penetrating the U.S. market, causing concern among some industry experts.
We are witnessing a classic "arms race" scenario in the AI sector, where each side tries to protect its own technologies while simultaneously gaining access to those of others. Alibaba's ban on Claude Code is not just a corporate decision but a symptom of a deep geopolitical divide that will only deepen, fragmenting the global AI market.