Over 80,000 Relay Servers Conceal Chinese Users of US Frontier AI
A network of more than 80,000 relay servers is enabling users in China to hide their identities while accessing advanced large language models hosted in the US. The activity is likely aimed at cloning the models for unauthorized use. The scale of the relay infrastructure highlights a growing challenge for AI providers in controlling access.
The discovery of this relay network points to a sophisticated effort to bypass geographic restrictions on advanced AI systems. By routing traffic through tens of thousands of intermediary servers, users can obscure their true location and origin, making detection and blocking significantly more difficult for service providers.
The scale involved suggests a coordinated operation rather than isolated attempts. Such infrastructure would allow repeated, systematic access to frontier models, potentially enabling the extraction of capabilities for replication. For AI companies, this represents an escalating arms race between access control measures and increasingly elaborate evasion techniques.
This development could reshape how AI developers approach model security and distribution. If large-scale model cloning becomes feasible through such networks, it may accelerate concerns about intellectual property protection and the concentration of advanced AI capabilities. Governments and enterprises relying on these models could face heightened risks of unauthorized replication, potentially influencing future deployment strategies and international technology policies.