White House Accuses China of 'Industrial-Scale' AI Theft as Anthropic Uncovers 24,000 Fraudulent Accounts Linked to Chinese Labs
Summary
The White House accuses China of 'industrial-scale' AI theft as Anthropic exposes roughly 24,000 fraudulent accounts tied to Chinese labs DeepSeek, MiniMax, and Moonshot AI, which generated over 16 million exchanges with Claude to steal proprietary data, while Congress pushes new sanctions legislation ahead of a high-stakes Trump-Xi summit where AI protection is becoming a critical national security flashpoint.
Key Points
- The White House Office of Science and Technology Policy accuses China of conducting 'industrial-scale' distillation campaigns to steal American AI capabilities, committing to share intelligence with OpenAI, Anthropic, and Google to strengthen defenses against foreign model theft.
- Anthropic reveals it has identified approximately 24,000 fraudulent accounts linked to three Chinese AI labs — DeepSeek, MiniMax, and Moonshot AI — generating over 16 million exchanges with its Claude model using jailbreaking techniques and proxy services to extract proprietary information.
- Congress is advancing the Deterring American AI Model Theft Act to impose sanctions on entities using unauthorized query-and-copy techniques, while the policy memo arrives three weeks before a Trump-Xi summit where AI model protection is emerging as a key national security and diplomatic negotiating point.