Anthropic is enforcing strict terms of service to prevent Chinese companies from accessing Claude Code, yet firms such as Ant Financial and ByteDance continue to bypass these restrictions using cloud services, overseas subsidiaries in Singapore, or virtual private networks. This situation is complicated by a separate ban imposed by Alibaba on its own employees, who are now required to delete all Claude models from their devices. The move follows reports of hidden code within the software that could flag users based in China or linked to a Chinese laboratory.
Anthropic’s Thariq Shihipar described the flagged code as an experiment conducted in March to stop account abuse and distillation, noting that stronger safeguards have since replaced it. The company has previously accused Alibaba, DeepSeek, Moonshot AI, and MiniMax of distilling their own smaller models using Claude outputs. While the ban targets potential intellectual property theft, the technical workarounds used by Chinese tech giants highlight the difficulty of enforcing software restrictions across borders.
* Hidden code was used to flag users in China or linked to Chinese labs
* Alibaba requires staff to delete all Claude models from their devices
* Bypass methods include cloud services, Singapore subsidiaries, and VPNs
