Anthropic accuses Alibaba's Qwen of stealing model via distillation programs
Alibaba's Qwen lab allegedly copied Anthropic's models over three months using distillation.
Anthropic has reportedly accused Alibaba's Qwen AI Lab of using 'distillation programs' to copy its proprietary AI models between April and June 2026. Distillation is a technique where a smaller model learns from a larger model's outputs, effectively allowing unauthorized replication. Anthropic claims this violates its intellectual property and has called on the U.S. government to act.
In response, Anthropic is urging stronger export controls, legal safeguards for model IP, and penalties for entities that engage in such practices. This incident underscores the escalating battle over AI model ownership and the risks of cross-border IP theft, particularly between U.S. and Chinese AI labs. It could set a precedent for how model distillation is treated legally.
- Anthropic claims Alibaba's Qwen lab used distillation programs to train on Anthropic's models from April to June 2026.
- Anthropic is urging the U.S. government to impose penalties and strengthen export controls on AI model theft.
- This case may set a precedent for legal treatment of model distillation as IP infringement.
Why It Matters
This could redefine IP protection in AI and escalate U.S.-China tech tensions over model theft.