U.S.-based artificial intelligence company Anthropic reportedly uncovered coordinated efforts by three Chinese AI firms to extract capabilities from its Claude chatbot at scale, describing the activity as industrial-level intellectual property theft.

The company alleged that DeepSeek, Moonshot AI, and MiniMax used a technique known as “distillation” — leveraging outputs from a more advanced AI system to improve the performance of a less capable one.

“These campaigns are growing in intensity and sophistication,” Anthropic said in a statement. “The window to act is narrow.”

Distillation is widely used in AI development to create smaller, more cost-efficient versions of models. However, Anthropic said the companies carried out the practice by generating roughly 16 million interactions with Claude through about 24,000 fake accounts.

According to the company, MiniMax conducted the largest operation, accounting for more than 13 million exchanges. The activity focused heavily on coding, agentic reasoning, and tool use — areas where Claude is considered particularly strong.

Anthropic argued that such efforts allow competitors to replicate advanced capabilities at a fraction of the development cost, potentially sidestepping U.S. export controls designed to protect sensitive AI technologies.

The company also warned that models built through unauthorized distillation may lack key safety guardrails, including restrictions intended to prevent assistance with bioweapons development or cyberattacks, raising potential national security concerns.