China’s Ministry of State Security is warning citizens against using foreign AI models, citing significant data privacy risks and potential cyber espionage threats from unauthorized third-party proxy services.

Key Points

• The Ministry of State Security (MSS) claims that accessing US-based AI models through grey-market "transfer stations" exposes users to data theft and backdoors.
• Research from the Oxford China Policy Lab identifies a thriving ecosystem on platforms like Taobao and Telegram reselling access to Anthropic’s Claude at discounted rates.
• Many illicit proxy services rely on bulk-registered accounts, stolen credit card data, and shared credentials, creating substantial financial and security risks for end users.
• Despite geopolitical tensions, US consumers continue to adopt Chinese models like DeepSeek V4 Pro and Alibaba’s Qwen 3.6 due to their lower costs and open-source flexibility.
• China is aggressively restructuring its domestic semiconductor and memory sectors to reduce reliance on foreign hardware, including high-end chips from Nvidia and AMD.

Why it Matters

This escalating rhetoric highlights a deepening technological decoupling as both nations attempt to secure their AI ecosystems against foreign influence and espionage. The trend suggests that users on both sides of the Pacific may soon face stricter regulatory barriers, potentially limiting access to global AI innovation in favor of national security.