Anthropic Accuses Chinese AI Labs of Data Extraction
- Three Chinese AI labs, DeepSeek, Moonshot, and MiniMax, allegedly extracted over 16 million responses from Anthropic’s Claude chatbot using approximately 24,000 fraudulent accounts.
- Anthropic claims these actions violate its terms of service and undermine U.S. export controls by stripping safety safeguards from the model.
- Critics on X accuse Anthropic of hypocrisy due to the common practice of training AI models on publicly available data.
The accusations have sparked a debate over intellectual property rights and the ethics of AI model training methodologies. Anthropic emphasizes the need for coordinated industry action to address large-scale distillation attacks.
The extraction activities by the Chinese labs involved millions of interactions with Claude’s chatbot, highlighting significant concerns about security and international collaboration in AI development. (Source)
