
Anthropic Moves to Close Loopholes Chinese Firms Use to Access Claude
FT reporting reveals Ant Group ran corporate Claude accounts through a Singapore entity while ByteDance reimbursed engineers for VPN-accessed subscriptions — practices Anthropic is now shutting down.
Anthropic is stepping up enforcement against Chinese companies that have been routing around its China ban, after Financial Times reporting on July 3 detailed the workarounds used by some of the country's biggest tech firms.
According to the FT, Ant Group provided employees with corporate Claude accounts linked to its Singapore-based entity, while ByteDance reimbursed engineers for personal Claude subscriptions they accessed over VPNs. The practices violate Anthropic's terms of service — which prohibit use by Chinese companies and foreign entities under their control — though not US or Chinese law.
Transfer Stations and Time Zones
Anthropic now plans to monitor accounts for signals such as computer time zones and usage patterns, and to target "transfer station" services that relay requests from China through overseas Claude accounts. The company has already detected and shut down accounts tied to offshore subsidiaries of restricted firms.
The crackdown carries a real financial cost. CEO Dario Amodei said in February that Anthropic had "forgone several hundred million dollars in revenue" by cutting off firms linked to the Chinese Communist Party and blocking state-sponsored actors who tried to abuse the system.
The Demand Problem
The loopholes underscore an awkward reality: Claude remains in heavy demand inside China's biggest AI companies even as domestic models close the gap on benchmarks. Engineers at firms shipping their own frontier models were, until this week, expensing subscriptions to a competitor their government cannot buy from.
Anthropic recently accused Alibaba of using Claude outputs for model distillation, and the enforcement push follows the discovery — and removal — of covert tracking code in Claude Code that flagged China-linked usage, an episode that triggered its own backlash and prompted Alibaba to ban the tool internally.
Between Two Governments
The move positions Anthropic ever more firmly on one side of the AI cold war. Having just emerged from a 19-day US export-control suspension of Fable 5, the company is demonstrating to Washington that its access controls have teeth — while Chinese firms accelerate their migration to domestic alternatives like DeepSeek, GLM and Qwen. The question for Beijing's tech giants is no longer whether they can keep borrowing American frontier models, but how fast their own can catch up.
Newsletter
Get Lanceum in your inbox
Weekly insights on AI and technology in Asia.

