China warns of security risks in AI relay services for foreign models

China's Ministry of State Security (MSS) issued a stark warning on Monday about the security risks of so-called "AI relay services"—intermediary platforms that aggregate access to both domestic and overseas large language models through a single interface. While these services offer convenience, lower prices, and access to leading US models like Anthropic's Claude and Microsoft-backed OpenAI's GPT series (neither officially available in mainland China), the MSS cautioned that some operators lack proper qualifications and employ weak security controls. This creates a thriving grey market where sensitive user data may be retained without adequate encryption, increasing the risk of privacy leaks and illicit data trading.

The warning underscores the tension between Chinese developers' demand for advanced foreign AI—particularly for stronger coding capabilities—and Beijing's stringent data sovereignty laws. Despite rapid progress by Chinese AI startups, many developers continue to rely on relay platforms to bypass restrictions. The MSS highlighted that such services operate without official oversight, potentially enabling unauthorized cross-border data transfers and exposing proprietary information. As AI relay markets expand, both developers and enterprises face heightened exposure to security breaches and regulatory scrutiny, making this a critical issue for tech professionals operating in or with China.