China's AI Distillation Strategy: The West Is Losing IP Faster Than It Can Innovate
Major U.S. AI companies including OpenAI, Google, and Anthropic are sharing intelligence about Chinese firms allegedly using 'distillation' techniques to extract capabilities from American AI models, with Anthropic specifically blocking Chinese-controlled companies from using Claude and identifying three Chinese AI labs—DeepSeek, Moonshot, and MiniMax—as illicitly extracting model capabilities, with the practice involving large-scale data requests to extract and reverse-engineer AI model capabilities.
But here's what's being undersold: it works. The last year shaped up as a big one for Chinese open-source models, with DeepSeek releasing R1, its open-source reasoning model, in January, shocking the world with what a relatively small firm in China could do with limited resources.
Chinese AI firms' near-unanimous embrace of open source has earned them goodwill in the global AI community and a long-term trust advantage, with more Silicon Valley apps expected to quietly ship on top of Chinese open models in 2026, with the lag between Chinese releases and the Western frontier expected to keep shrinking from months to weeks and sometimes less.
The geopolitical irony is devastating: US companies are trying to restrict Chinese access to their models to preserve competitive advantage. But Chinese labs are building better open models and gaining developer loyalty in the process. Western closedness = Chinese openness advantage.
My take: The US is winning the frontier AI race (barely). China is winning the developer ecosystem war (decisively). By 2027, the gap between cutting-edge Western models and reproducible Chinese models will be negligible. The IP theft argument is valid, but it's also a distraction from the fact that US companies have made openness a competitive liability.