Three Chinese AI labs (DeepSeek, Moonshot, and MiniMax) conducted large-scale distillation attacks against Anthropic's Claude model, using over 24,000 fraudulent accounts to generate 16+ million exchanges and extract capabilities for their own models.
Anthropic detected industrial-scale campaigns by three Chinese AI laboratories - DeepSeek, Moonshot, and MiniMax - to illicitly extract Claude's capabilities through a technique called 'distillation.' The labs generated over 16 million exchanges with Claude through approximately 24,000 fraudulent accounts, violating terms of service and regional access restrictions. DeepSeek generated over 150,000 exchanges targeting reasoning capabilities and censorship-safe alternatives. Moonshot AI conducted over 3.4 million exchanges focusing on agentic reasoning, tool use, and computer vision. MiniMax executed over 13 million exchanges targeting agentic coding and tool orchestration. The labs used proxy services and 'hydra cluster' architectures to distribute traffic and evade detection. Anthropic attributed these campaigns through IP address correlation, request metadata, and infrastructure indicators. The attacks targeted Claude's most differentiated capabilities including agentic reasoning, tool use, and coding. Anthropic warns that illicitly distilled models lack safety guardrails and could enable authoritarian governments to deploy frontier AI for offensive cyber operations, disinformation campaigns, and mass surveillance.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Vulnerabilities that can be exploited in AI systems, software development toolchains, and hardware, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
No population impact data reported.
