Anthropic Accuses Chinese AI Firms of 'Industrial-Scale' Model Theft

Key Intelligence

• ■Anthropic claims Chinese AI companies DeepSeek, MiniMax, and Moonshot engaged in 'distillation attacks' to illegally extract data from its Claude AI model.
• ■The alleged theft involved using 24,000 fraudulent accounts and conducting approximately 16 million interactions to train their own smaller AI models.
• ■Anthropic is developing new detection and prevention techniques to combat such 'model ripping' and is urging the broader AI industry to address model theft.
• ■The accusations highlight significant intellectual property and data security challenges within the rapidly developing global AI sector.