Anthropic catches Chinese labs copying Claude

Anthropic just revealed that DeepSeek, Moonshot, and MiniMax ran coordinated ops to siphon Claude's capabilities through 16M+ fraudulent exchanges across 24K fake accounts a threat it says demands immediate industry-wide action.

Anthropic accused the labs of distillation via the conversations, which involves training a weaker system on the outputs from a stronger one.

MiniMax ran the largest campaign at 13M+ exchanges, and Anthropic caught it mid-operation, spotting the lab shift focus to a new release in under 24 hours.

DeepSeek had Claude spell out reasoning step-by-step and rewrite politically sensitive queries, generating training data for both logic and censorship.

OAI raised similar concerns with Congress weeks ago, and Anthropic is now pushing for coordinated action from the broader AI industry and gAnthropic catches Chinese labs copying Claude