Anthropic’s China geofence is becoming a cat-and-mouse story
A Wired report says people in China are finding ways around Anthropic’s geolocation restrictions.
AI access controls are becoming a live operational problem, not merely a policy line.
Wired reports on how people in China keep outsmarting Anthropic’s geolocation restrictions. The source item does not detail the methods or the scale of the workarounds. It does establish the central dynamic: a frontier AI provider is trying to restrict access by geography, while users are finding ways around the boundary.
That is the problem with software borders. They can be imposed quickly, but they must be defended continuously. Every restriction creates an incentive for routing, resale, account sharing or other forms of arbitrage.
The story also complicates the public debate over AI controls. Policy can declare that a model should not be available in a jurisdiction. The internet then asks a narrower question: can the restriction actually hold?
For AI companies, the answer matters commercially and politically. A geofence that leaks invites scrutiny from governments. A geofence that is too aggressive risks cutting off legitimate users. The Anthropic case, as reported by Wired, sits squarely in that tension.