China Warns of Security Risks in Foreign AI Model Relay Services
- tech360.tv
- 2 hours ago
- 2 min read
China's national security authority has warned of risks in using artificial intelligence relay services that provide access to overseas AI models. The Ministry of State Security described such services as intermediaries that aggregate access to multiple domestic and overseas models through a single interface.
The ministry highlighted concerns over data leaks, privacy breaches, and unauthorised cross-border data transfers. These issues occur amid a thriving grey market for restricted foreign systems.
While relay platforms offer convenience, lower prices, and access to foreign models, some operators pose data security risks. The ministry noted that some platforms operate without proper qualification and with weak security controls.
Chinese developers turn to relay services to access leading US AI models such as Anthropic's Claude or Microsoft-backed OpenAI's GPT series. Neither of these companies officially supports users in mainland China.
Some relay platforms retain user data on their servers without adequate encryption. This practice creates risks that sensitive information could be leaked or sold.
Other concerns include backdoors implemented in some AI relay platforms. These hidden methods of accessing a computer system or network could be exploited to steal accounts and surveil developer devices.
Some operators substitute premium models with cheaper alternatives. This practice misleads clients while reducing performance.
Chinese developers continue to seek US models for their stronger coding capabilities. This preference persists despite rapid advances by Chinese AI start-ups, including DeepSeek and Zhipu AI.
US systems topped the Coding Index ranked by AI benchmark platform Artificial Analysis. Models from OpenAI, Anthropic, and Google took up the top six spots, including GPT-5.5, GPT-5.4 Mini, Claude Opus 4.8, Claude 4.7, and Gemini 3.1 Pro Preview.
The highest-ranking Chinese model was Alibaba Group Holding's Qwen 3.7 Max, followed by DeepSeek's V4 Pro. Alibaba owns the South China Morning Post.
The White House warned that Chinese entities were conducting large-scale efforts to access advanced US AI systems through networks of intermediary accounts. Anthropic previously disclosed attempts by China-linked actors to gain access to its models through coordinated proxy infrastructure.
The Ministry of State Security advised local developers to choose authorised platforms and take precautions to protect user data.
China's Ministry of State Security warned that grey-market AI relay platforms pose risks of data leaks, privacy breaches, and unauthorised data transfers.
Many Chinese developers use these unregulated services to access US models like OpenAI's GPT series and Anthropic's Claude, which do not officially operate in mainland China.
Security officials cautioned that some platforms lack proper qualifications, lack adequate encryption, substitute premium models with cheaper alternatives, and contain backdoors that could expose developer devices.
Source: SCMP
