Chinese authorities have launched investigations into AI companies DeepSeek and Moonshot AI. The probes follow allegations that these firms funneled sensitive user data to US servers by using Anthropic’s Claude models through proxy networks, violating terms that prohibit model distillation.
Anthropic claims that Chinese firms created approximately 24 000 fraudulent accounts to conduct over 16 million exchanges with Claude. These actions were allegedly aimed at reverse-engineering the model's reasoning and coding capabilities to reduce development costs.
Data indicates that DeepSeek conducted over 12 million exchanges in 14 days in July 2026. Moonshot AI reportedly routed 300 000 customer requests through 5 380 accounts in September 2026, in addition to 23 million exchanges between May and July. Another firm, MiniMax, was linked to 13 million exchanges.
In September 2026, the NSA, CISA, and FBI issued a joint advisory regarding a network of six Chinese companies engaged in industrial-scale distillation of US AI models. The agencies noted these operations have been active since late 2024.
