Relays Are Masking Chinese Access to Frontier AI Models in the US

A recent report from Dark Reading highlights a significant security concern involving the use of over 80,000 AI relay servers. These servers are being utilized by users in China to mask their identities while accessing advanced large language models (LLMs) hosted in the United States. Security researchers suggest that this infrastructure is likely being deployed to facilitate the unauthorized cloning of proprietary frontier AI models. By routing traffic through these relay networks, actors can bypass geographic restrictions and security protocols designed to prevent the export of sensitive AI technology. This development underscores the growing challenge of securing high-value AI assets against sophisticated evasion techniques. As frontier models become increasingly central to national and economic security, the ability to monitor and control access to these systems remains a critical priority for cybersecurity professionals and policymakers alike.