The Allegations
Anthropic has formally accused Alibaba and its Qwen AI lab of conducting what it describes as the largest known distillation attack against its Claude models. The accusations were made in a letter to the U.S. Senate, first reported in June 2026.
Scale of the Attack
According to Anthropic's investigation:
- Duration: April 22 to June 5, 2026 (approximately 6 weeks)
- Fraudulent accounts: Nearly 25,000
- Total exchanges: Approximately 28.8 million
- Method: Commercial proxies to disguise the origin of requests
What Is Distillation?
Distillation (or model distillation) is a technique where a smaller, less capable model is trained using outputs from a larger, more capable model. By using Claude to generate millions of responses and training Qwen models on those outputs, Alibaba could effectively harvest American AI capabilities at a fraction of the normal development cost.
Industry Context
The allegations come amid heightened U.S.-China tensions over AI technology and potential national security implications of frontier AI model development.
Alibaba's Response
Alibaba has not directly addressed the specific allegations. The company has historically denied engaging in IP theft, and Qwen has emerged as a capable open-source AI series competing with Western frontier models.
Regulatory Implications
The incident is likely to intensify calls for:
- Export controls on advanced AI models
- Mandatory compute thresholds for frontier model training
- International norms around AI development practices
Anthropic's Position
Anthropic has been vocal about the need for stronger protections around frontier AI models, citing this incident as evidence that self-regulation is insufficient.