Alibaba to ban employees from using Anthropic’s coding tool, source says
Chinese technology company Alibaba has reportedly prohibited employees from using Anthropic’s Claude Code for work-related tasks after the AI coding assistant came under scrutiny over features capable of identifying users with links to China.
According to a source familiar with the matter, the directive forms part of an escalating dispute between Alibaba and the U.S.-based artificial intelligence company, following allegations by Anthropic that Alibaba improperly sought to replicate the capabilities of its Claude AI model.
Claude Code, Anthropic’s AI-powered coding assistant for software developers, has gained popularity among programmers in China despite restrictions limiting access to users and organisations in the country.
The source, who was not authorised to speak publicly, said Alibaba has instructed staff to use the company’s in-house coding platform, Qoder, instead of Claude Code.
Neither Alibaba nor Anthropic immediately responded to requests for comment. Alibaba has also not publicly addressed Anthropic’s allegations.
Last month, Anthropic claimed it had been targeted by what it described as a “distillation” attempt by Alibaba—a process in which a smaller AI model is trained using the outputs of a more advanced model to reproduce similar capabilities.
In a letter reportedly sent to two U.S. senators and reviewed by Reuters, Anthropic said the alleged effort was aimed at accelerating the development of AI systems capable of matching the performance of its advanced Mythos Preview model.
Alibaba’s reported decision comes days after software developers raised concerns that Claude Code contained features designed to inspect users’ operating environments, including details such as time zones and proxy settings, while embedding hidden markers into prompts transmitted to Anthropic’s servers.
An Anthropic employee later explained on X that the feature had been introduced as an experimental measure in March to combat account abuse by unauthorised resellers and reduce the risk of model distillation.
The source said that while Anthropic’s measures targeting users in China can often be bypassed by individuals using servers located in the United States, companies are generally more cautious because of legal and regulatory compliance considerations.
As American AI firms tighten controls to prevent unauthorised access, resale and replication of their technologies, Chinese cloud computing and artificial intelligence companies have increasingly turned to locally developed and open-source models, including DeepSeek, Alibaba’s Qwen, Moonshot and Zhipu.
At the same time, Chinese AI models are gaining traction in the United States, a trend that has raised concerns among some industry observers over growing global competition in artificial intelligence.
Chinese media first reported Alibaba’s internal directive.
About The Author
