Y Combinator’s Garry Tan wants U.S. open-weight AI labs to ‘distill’ frontier models, too
techcrunch ·
Garry Tan, CEO of Y Combinator, said he would prefer regulators not to stop Chinese AI labs from distilling knowledge from frontier models, and suggested U.S. open-weight labs could use similar methods. In an interview with CNBC, he said he would do nothing and raised the idea of an American distillation regime. He told TechCrunch he wants smaller U.S. open-weight AI labs to train on American frontier AI labs, creating more robust open-weight alternatives not originating from China. Distillation is described as extensively prompting another model to learn how it reasons and is a common, legitimate technique. Anthropic released a second report this week alleging Chinese labs conduct illicit distillation attacks, using hidden identities, fraud, and stolen credentials, and CEO Dario Amodei has called for a regulatory crackdown; Tan disagrees. Tan does not advocate using stolen credentials, saying U.S. labs should be allowed to come through the front door. He argues it is overreach for AI companies to dictate customer use of information their models provide, and notes proprietary labs scraped human knowledge and copyrighted material without permission. He also wants balance: frontier labs should remain fundable and viable, while open-weight models provide freedom and access, and he sees the nightmare scenario as one monolithic company controlling AI.
techcrunch