China's Cyberspace Administration (CAC) is investigating DeepSeek and Moonshot AI (maker of the Kimi assistant) over allegations they routed real requests from Chinese users to Claude, the model built by US-based Anthropic, without those users' knowledge. It's a story with a striking twist: the usual regulatory worry in AI is data flowing OUT of the West and into China, not the other way around.
According to reports, the case started on Anthropic's side: the company was monitoring attempts by rival companies to use Claude to train or improve competing models (a "distillation" practice several labs actively watch for), and in that process it identified large-scale activity linked to DeepSeek and Moonshot. CAC officials have already visited both companies' offices to interview executives and employees.
The most sensitive part of the case: according to Anthropic, in at least some instances Moonshot forwarded genuine requests made by real Kimi users to Claude for processing — meaning people using what they believed was a 100% Chinese assistant unknowingly ended up with their query processed by a US AI system.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel