Y Combinator's Tan Says Let US Labs Distill Frontier Models

Y Combinator CEO Garry Tan says the U.S. should let its own open-weight AI labs distill frontier models the same way Chinese labs have been accused of doing, an argument he made to CNBC on Sept. 10 and to TechCrunch on Sept. 11.

Distillation trains a smaller model to copy a larger one's outputs, letting the smaller model approach the bigger model's performance at a fraction of the compute cost. It is also how much of the current wave of cheap, open-weight Chinese models were reportedly built, by querying closed frontier systems and training on the results.

What Tan told CNBC

Asked what regulators should do about labs distilling from proprietary models without permission, Tan told CNBC, in remarks the outlet published Sept. 10 at 9:37 p.m. Eastern: "I would do nothing." He went further: "We could argue that there should be an American distillation regime." Pressed on what he would target instead, he said: "If I were a regulator, that's what I would go after," pointing at restrictions on American labs rather than at the Chinese labs doing the copying. Tan named OpenAI, Anthropic, Moonshot AI, DeepSeek and MiniMax in the same interview, arguing that American open-weight labs need the same freedom to learn from frontier systems that Chinese labs have taken. "We need to be focused on science fact, not science fiction," he said, dismissing fears that distillation alone hands China a lasting edge.

The line he draws with TechCrunch

A day later, Tan told TechCrunch: "Controlling what users and customers do with API calls to closed weight models feels constraining." He was careful to distinguish his proposal from the fraud allegations aimed at Chinese labs, telling the outlet he is not advocating for American labs to use stolen credentials, only for them to be free to access frontier models' APIs legitimately. "The nightmare scenario, the doomer scenario for AI, is that there's just one company," he said.

Date	Outlet	What Tan said
Sept. 10, 2026	CNBC	"I would do nothing"
Sept. 11, 2026	TechCrunch	"Controlling what users and customers do with API calls... feels constraining"

Amodei's counter-argument

Tan's position runs directly against Anthropic's. TechCrunch reported that Anthropic CEO Dario Amodei has previously called on U.S. regulators to crack down on distillation, and that Anthropic released a report alleging Chinese labs run "illicit distillation attacks," hiding their identities and relying on fraud and stolen credentials to copy frontier models without permission. Tan's framing accepts that Chinese labs have been distilling U.S. models; his disagreement with Anthropic is about what follows from that fact.

Why he frames it as a monopoly question

Tan told CNBC he wants frontier labs like OpenAI and Anthropic to stay "fundable" and keep "a great business model," while also wanting open-weight models to give smaller labs and developers freedom and access. Asked about the pace of change, he told CNBC it will "take decades for this to actually percolate into society," calling that a good thing rather than a reason to slow down. He did not say whether he has raised the distillation-regime idea with any lawmaker or agency.