Chinese AI Giants Accused of Sending Millions of User Queries to U.S. Models

Anthropic alleges that Chinese AI firms DeepSeek and Moonshot funneled millions of real user queries through thousands of fake accounts to distill and replicate its models. If accurate, it raises both competitive questions about how quickly frontier capabilities can be copied and privacy questions about where users' prompts actually end up. It also foreshadows tougher access controls and a new front in US-China AI tensions.

• Anthropic accuses DeepSeek and Moonshot of using thousands of fake accounts to extract model capabilities via distillation.
• Millions of real user queries were allegedly routed to U.S. models, raising user privacy and data-handling concerns.
• The claims sharpen scrutiny of model distillation and could push AI providers toward stricter API access enforcement.