Anthropic Accuses Chinese Firms of Using Claude for AI Training

Anthropic has detailed sweeping allegations of industrial-scale intellectual property theft. The company asserts that Chinese AI firms, including DeepSeek, orchestrated massive campaigns to misuse its Claude AI for training their own models. The alleged operation was vast in scale, involving the creation of roughly 24,000 fraudulent user accounts that engaged in over 16 million exchanges with Claude's API. This volume of interaction suggests a systematic, coordinated effort to generate a huge dataset of Claude's inputs and outputs. Such a dataset would be incredibly valuable for a technique called distillation, where a smaller, less expensive model is trained to mimic the behavior and capabilities of a larger, more advanced one like Claude. This represents one of the most significant public accusations of AI model extraction to date. It goes beyond simple terms-of-service violation, pointing to a sophisticated effort to replicate proprietary AI advancements. The allegations underscor