Y Combinator's Garry Tan advocates for a US-based AI model distillation regime

Y Combinator CEO Garry Tan has suggested that U.S. AI labs should consider implementing an American distillation regime, allowing smaller open-weight labs to learn from frontier models. Distillation, a process where one model prompts another to understand its reasoning, is commonly used in the industry. Tan’s comments contrast with recent reports from Anthropic, which accused Chinese labs of engaging in "illicit distillation attacks" by hiding their identities and using stolen credentials. Anthropic CEO Dario Amodei has previously called on U.S. regulators to crack down on such activities. Tan argues that it is an overreach for AI labs to dictate how customers use the information shared by their models. He also points out the irony that proprietary labs did not seek permission when ingesting vast amounts of copyrighted human knowledge to train their own models. He suggests that rather than banning the practice, there is a role for the government to normalize how users and customers interact with closed-weight model APIs.