Alibaba has told its employees to stop using Anthropic’s Claude Code at work, according to a person familiar with the order. The ban comes after the AI coding tool drew scrutiny for features that can identify China-linked users, raising compliance concerns inside the Chinese tech giant.
The move is part of a growing conflict between Alibaba and Anthropic. Last month, Anthropic accused Alibaba of illicitly extracting capabilities from its Claude AI model through a process called “distillation,” where a weaker model is trained on the outputs of a more powerful one. Alibaba has not publicly responded to those accusations, and neither company replied to requests for comment.
The person who spoke about the ban was not authorized to talk to media and asked not to be identified. They said Alibaba employees are now being directed to use the company’s own coding platform, Qoder, instead.
Claude Code is Anthropic’s AI assistant built for software developers. Despite Anthropic restricting access for users and companies in China, the tool has remained popular among Chinese programmers. That popularity is part of what makes enforcement so difficult. As the source explained, individual users can route traffic through U.S.-based servers to make it appear as though requests are coming from inside the country. Companies, on the other hand, are more exposed to legal and compliance risks, which makes a formal workplace ban both more enforceable and more significant.
The timing matters. Just days before Alibaba issued its internal order, developers flagged that Claude Code contained features that inspected user environments, including timezone data and proxy-related information, and inserted subtle markers into prompts sent back to Anthropic’s servers. An Anthropic employee confirmed on X that this was “an experiment we launched in March” designed to detect account abuse by unauthorized resellers and to protect against model distillation. Critics read it differently, describing it as a surveillance mechanism aimed at Chinese users.
Anthropic says the distillation effort by Alibaba was specifically targeting its most advanced model, Mythos Preview, and that the attack could help China accelerate its AI development. The company detailed this in a letter sent to two U.S. senators, a copy of which was seen by Reuters.
This dispute sits inside a much larger story. The U.S. and China are in a serious race to lead in artificial intelligence, and American AI companies have grown increasingly aggressive about protecting their models from being copied or reverse-engineered. The response from Chinese firms has been to lean harder into domestic alternatives:
- DeepSeek, the open-source model that stunned Western observers earlier this year
- Alibaba’s own Qwen model family
- Moonshot and Zhipu, two other Chinese AI startups building independent capabilities
At the same time, Chinese AI models are starting to gain traction in the U.S. market, which has alarmed some American industry experts who worry about the security implications of that shift.
Alibaba’s ban was first reported by Chinese media outlets before Reuters confirmed the details independently. Whether other Chinese companies follow with similar restrictions on foreign AI tools remains to be seen, but the direction of travel is clear. As trust between U.S. and Chinese tech firms erodes, the tools employees are allowed to use are becoming another front in a much bigger standoff.
