How Chinese AI Firms Tried to Clone U.S. AI Models

Anthropic alleges that Chinese AI companies DeepSeek and Moonshot tried to replicate its models through 'distillation,' using thousands of fake accounts and millions of queries to harvest outputs. If accurate, it shows how hard it is for leading labs to protect their most valuable asset — model behavior — and raises the stakes in US-China AI competition.

• Anthropic accuses DeepSeek and Moonshot of distillation-based cloning of its AI capabilities.
• The alleged method relied on thousands of fake accounts and millions of real user queries.
• Model outputs are a leak vector, pushing labs toward tighter access controls and abuse detection.