Anthropic is tightening enforcement to stop Chinese companies from using its Claude AI models through indirect routes, according to reporting from the Financial Times.
The FT reports that despite existing restrictions on Chinese access, engineers kept finding ways in — reaching Claude through overseas subsidiaries, third-party cloud providers, and VPNs. Sources cited by the FT name Chinese firms including Ant Group among those that tapped the models via these workarounds. Anthropic is now moving to close those loopholes.
Firstpost reports the crackdown follows accounts of Chinese firms accessing Claude through foreign subsidiaries, cloud services, and VPNs, and describes Anthropic as shutting those secret routes down.
Running in parallel is a move from the other direction. Multiple outlets, including NDTV and Pandaily, report that Alibaba is banning Claude Code — Anthropic's coding tool — for its employees, citing security concerns. According to Pandaily, Alibaba has ordered staff to remove all Anthropic products, with the ban taking effect July 10. Some coverage, including The News International, ties Alibaba's decision to "backdoor" claims.
The two developments point in the same direction: a widening separation between American AI providers and Chinese tech giants. Anthropic bars Chinese firms from its models on one side, while a major Chinese company pushes those same tools out of its own workplace on the other.
Why it matters: The episode shows that AI access is becoming a front line in US-China technology tensions, where policy on paper is only as strong as the workarounds — cloud accounts, offshore units, VPNs — that companies use to get around it.