Anthropic Moves to Close Loopholes Giving Chinese Firms Access to Claude Code
Anthropic is closing the workarounds that let Chinese tech firms keep using Claude Code, according to Financial Times reporting summarized by The Decoder on July 3. Per the reporting, firms including Ant Financial and ByteDance have gotten around the restrictions through cloud services, overseas subsidiaries in Singapore, or VPNs.
Anthropic's terms of service explicitly ban sales to companies controlled by China, and the company has previously accused Alibaba, DeepSeek, Moonshot AI, and MiniMax of using Claude outputs to train competing models through distillation. The enforcement is now bidirectional: Alibaba has banned its own employees from using Claude Code and ordered deletion of Claude models, after discovering embedded monitoring code that flagged China-based users. Anthropic's Thariq Shihipar described that code as a March experiment against account abuse, since replaced by stronger safeguards.
For European enterprises the relevant lesson is about enforcement mechanics. Geographic restrictions on frontier models are moving from paper terms to active detection, identity verification, and usage-pattern analysis. Any multinational routing AI access through subsidiaries should assume the providers are now looking.