Y Combinator chief executive Garry Tan wants Washington to leave distillation alone, and American open-weight labs to try it on domestic frontier models.
“I would do nothing,” Tan told CNBC, adding that “we could argue that there should be an American distillation regime.” He later elaborated to TechCrunch that smaller US open-weight labs should be free to train on output from American frontier labs, giving builders a stronger non-Chinese option.
Distillation is the practice of extensively prompting one model to learn how it reasons. Frontier labs treat it as legitimate internally but object when outside parties do it without permission. Anthropic this week published a second report alleging Chinese labs run what it calls illicit distillation attacks using fake identities and stolen credentials, and chief executive Dario Amodei has called for regulators to crack down.
Tan’s objection is twofold. He argues it overreaches for a lab to dictate what customers do with information its models hand over, and he notes the proprietary labs did not ask permission when they ingested vast amounts of human knowledge, including copyrighted material, to train their own systems.
He also wants a balance between closed frontier labs and open-weight challengers. In his telling, the real doomsday scenario is not open models but a single company holding the best researchers and capital, monopolizing frontier AI. Tan’s position puts the accelerator that funds hundreds of AI startups at odds with the labs those startups increasingly build on top of.
