Alibaba Bans Employees from Using Anthropic's Claude Code
Starting July 10, Alibaba will restrict its workforce from utilizing Claude Code, a programming tool developed by Anthropic, as reported by various sources. Anthropic has already prohibited Chinese firms and their international branches from accessing its models and has been working to close any loopholes that might allow Chinese users entry. A recent post on Reddit suggested that a version of Claude Code could inadvertently track Chinese users. Thariq Shihipar from Anthropic clarified on X that this was part of an experiment initiated in March aimed at curbing account misuse by unauthorized resellers and preventing distillation. He mentioned that the team has since reinforced safeguards and intends to discontinue the experiment. Alibaba has deemed Claude Code as high-risk and is advising employees to utilize its own Qoder tool instead.
Key facts
- Alibaba will ban employees from using Anthropic's Claude Code starting July 10.
- Anthropic prohibits Chinese companies and their foreign entities from using its models.
- Anthropic has been closing loopholes that allow Chinese users to access Claude.
- A version of Claude Code could secretly identify Chinese users.
- Thariq Shihipar said the identification was an experiment to prevent account abuse and distillation.
- Alibaba classified Claude Code as high-risk software.
- Alibaba instructs employees to use its own Qoder tool instead.
- Distillation is training AI models on outputs of other models.
Entities
Institutions
- Alibaba
- Anthropic
- Alibaba Group Holding
- South China Morning Post
Locations
- China