Y Combinator CEO Opposes U.S. Crackdown on AI Distillation, Backs American Open-Weight Labs

Model distillation uses outputs from a more capable AI model to train a smaller or less capable one. Anthropic has accused Chinese companies including Moonshot AI, DeepSeek, and MiniMax of the practice. OpenAI believes DeepSeek's V3 and R1 model architectures were distilled from its GPT-4 and GPT-4o models. CNBC reported that Silicon Valley giants and national security experts are calling for action against Chinese companies engaged in model distillation. On Tuesday, the U.S. National Security Agency, Cybersecurity and Infrastructure Security Agency, and Federal Bureau of Investigation released an official cybersecurity advisory warning on the topic.

Tan, however, said he would do nothing. "We could argue that there should be an American distillation regime," he said, according to Slashdot. He argued regulators should focus less on curbing distillation and more on creating an equilibrium between open-weight models and frontier models, provided frontier models retain a price premium that keeps their business model feasible. "This is actually the ideal case. You want open weight models to give people freedom and access," he said. "If I were a regulator, that's what I would go after."

He acknowledged the balance is difficult, calling it "a tightrope." He said it is worth pursuing and "could result in the best possible outcome."

Tan later told TechCrunch he wants to see more American open-weight options that are not Chinese, built by smaller U.S. open-weight AI labs using the same training techniques on products from American frontier AI labs. Anthropic CEO Dario Amodei had previously publicly called on U.S. regulators to crack down on distillation. The Y Combinator CEO disagrees.

Tan is not advocating that American AI labs use stolen credentials to distill, according to the report. He wants them to be free to come in the front door. His argument is twofold. He says it is an overreach for AI labs to dictate what their customers can do with the information their models share with them. He also notes that proprietary AI labs did not ask permission when they vacuumed up as much human knowledge as they could to train their models, and ingested copyrighted material without permission from intellectual property holders. "Controlling what users and customers do with API calls to closed weight models feels constraining, and there's a role government can play here to normalize the fact that access to intelligence that was trained on broad public access data should itself also be more a form of a public good than something locked away behind restrictive terms of service," he told TechCrunch.

Tan described the true AI doomer scenario as all the immense power of frontier AI winding up in the hands of a single powerful, proprietary provider. "The nightmare scenario, the doomer scenario for AI is that there's just one company," he said. "It has the best access to capital. It has the best AI researchers. It runs away with it and suddenly there's one company that's monolithic. And that would be bad."