Alibaba bans Claude Code for employees, pushes its own Qoder AI tool

Full text

Alibaba bans access to Claude Code, tells employees to use internal Qoder instead

Anthropic was tracking markers to indicate which users were in China

Anthropic accused Alibaba of major Claude model distillation effort

Alibaba has reportedly banned its employees from using Claude Code internally, beginning July 10 2026, classifying it as a high-risk tool that risks organizational security.

The change follows similar trends already observed among American tech giants, banning Chinese tools from internal use, but Alibaba cited genuine concerns that have been acknowledged by Claude-maker Anthropic.

The ban could also be seen as a push for Alibaba's own alternative, with workers advised to use the company's own Qoder AI assistant instead.

Alibaba bans Claude Code over security concerns

The controversy stems from developers reverse-engineering Claude Code, revealing it contained code to identify Chinese users. Checks for Chinese system time zones, proxy servers, AI lab infrastructure and network characteristics were all revealed.

Anthropic stated this experimental feature launched in March, and was designed to combat unauthorized resellers, prevent account abuse and protect its models from AI distillation.

However, the spyware was reportedly hidden using obfuscation and steganographic techniques, making them effectively invisible to users.

This isn't the first time both companies have found themselves in a sticky situation – Anthropic recently accused Alibaba of conducting the largest known model distillation attack against Claude (via Reuters ).

More broadly, Chinese companies have been increasingly referring to domestic AI tools like Qwen, DeepSeek, Moonshot and Ship amid growing geopolitical tensions.

While that trend has been largely mirrored in the US, in favor of the likes of OpenAI, Anthropic, Google Cloud and xAI, US firms have reportedly been exploring cheaper Chinese alternatives in the name of cost efficiency.

Alibaba and Anthropic haven't publicly commented on this matter as yet.