Key facts
- Anthropic accuses Alibaba Group of conducting a "distillation attack" on its Claude AI model.
- Anthropic identifies DeepSeek, Moonshot AI, and MiniMax in illicit AI capability extraction campaigns.
- Alibaba's alleged attack is described as the "largest known distillation attack" on Claude.
- The campaigns by DeepSeek, Moonshot AI, and MiniMax are characterized as industrial-scale.
- The accusations concern the illicit extraction of AI capabilities from Anthropic's Claude model.
US AI company Anthropic has accused several Chinese technology firms of illicitly extracting capabilities from its Claude AI model. The most prominent accusation is leveled against Alibaba Group, which Anthropic claims conducted "the largest known distillation attack" on Claude. This type of attack involves training a smaller, more efficient model to mimic the performance of a larger, more complex one, effectively stealing its capabilities.