Y Combinator's Tan advocates for American AI distillation of frontier models
Y Combinator CEO Garry Tan said U.S. open-weight AI labs should be allowed to use distillation techniques on frontier models, rather than regulators cracking down on Chinese labs. He argues that controlling API usage is overreach and that AI labs have already used public data without permission. Tan wants a balance between open-weight and frontier labs.
Tan’s stance directly counters Anthropic’s recent push for regulatory action against Chinese labs accused of illicit distillation, which allegedly involved fraud and stolen credentials. He distinguishes between such abuses and legitimate distillation, arguing that U.S. open-weight labs should be able to use the technique openly on frontier models. His reasoning draws on the fact that proprietary labs trained on vast amounts of public and copyrighted data without permission, so restricting API-based learning seems inconsistent. Tan also frames the issue as a competitive necessity, suggesting that allowing American distillation would create stronger domestic open-weight alternatives and prevent a monopoly-like concentration of AI power in a single closed provider.
This debate could shape how AI regulation evolves, affecting startups, researchers, and consumers. If Tan’s view gains traction, U.S. open-weight models may proliferate, offering more accessible alternatives to proprietary systems—but also raising questions about intellectual property and fair use. Conversely, stricter rules could protect frontier labs’ investments but risk centralizing AI power. Society may see a trade-off between innovation and control, with smaller players gaining leverage while larger firms face pressure on their business models. The outcome could influence who ultimately benefits from AI’s capabilities.