Garry Tan Calls for U.S. 'Distillation Regime' to Bolster Open-Weight AI
AI-generated image · Bay Street Wire
Y Combinator's CEO argues that American open-weight labs should be permitted to extract knowledge from frontier models to prevent a monolithic industry.
Y Combinator CEO Garry Tan is advocating for a "American distillation regime," suggesting that U.S. open-weight AI labs should be allowed to use distillation techniques to learn from American frontier AI models, according to TechCrunch.
Distillation involves a model maker using extensive prompting to learn the reasoning and functionality of another model. While common in AI training, the practice has become a point of contention. TechCrunch reports that Anthropic CEO Dario Amodei has called for regulatory crackdowns on the practice, and Anthropic recently released a report alleging that Chinese labs are conducting "illicit distillation attacks" using stolen credentials and fraud.
Tan told CNBC and TechCrunch that he opposes regulatory interference in this area. He argues that proprietary labs previously ingested vast amounts of copyrighted human knowledge without permission, and therefore, it is an overreach for those labs to restrict how customers use API calls. Tan suggests that intelligence trained on broad public data should function more as a public good than as a product locked behind restrictive terms of service.
Tan's push for distillation is framed as a necessity for competition. He told CNBC that while frontier labs drive the field forward and should remain fundable businesses, open-weight models provide essential freedom and access. Tan warned that the "doomer scenario" for the industry would be the emergence of a single, monolithic company that controls the best researchers and capital, effectively running away with the technology.