AI distillation, a technique where one AI model is trained using outputs from another, has evolved from a benign research idea into a shadow economy threatening the business model behind trillions of dollars in AI investment. US AI giants spend billions on data, talent, and computing power to build leading models, hoping to charge premium prices, but distillation allows rivals to quickly replicate those models cheaply, potentially eroding returns. "Generally, AI companies distill other AI companies," Elon Musk said during a legal battle with OpenAI this year. Anthropic accused Chinese tech giant Alibaba of using malicious distillation, such as creating tens of thousands of fake Claude accounts to harvest answers, with Sarah Heck, Anthropic's head of policy, writing that it "inverts the economic logic that underwrites American AI leadership." OpenAI warned distillation could create models that outperform today's frontier systems. AI stocks slumped after Chinese companies released new models, including GLM-5.2 from Z.ai, with some researchers believing it benefited from distilling US systems. Distillation began in 2015 when Google researchers described training a smaller model using its own large model, but after ChatGPT's 2022 release, it expanded to leveraging competitors' outputs.