Relays Are Masking Chinese Access to Frontier AI Models in the US
Summary
Over 80,000 AI relay servers are being used by individuals in China to obscure their identities while accessing advanced large language models (LLMs) in the US. This activity is suspected to be for the purpose of cloning these AI models.
IFF Assessment
The use of relays to mask access to AI models by actors in China suggests potential intellectual property theft and the illicit cloning of advanced AI technology, which poses a threat to the developers and the AI ecosystem.
Defender Context
This indicates a sophisticated effort to exfiltrate or replicate cutting-edge AI models, potentially for malicious purposes or competitive advantage. Defenders should be aware of the increasing use of intermediary servers to bypass access controls and monitor for unusual traffic patterns originating from known adversarial regions targeting AI infrastructure.