Y Combinator's Garry Tan opposes crackdown on AI distillation, advocates 'American distillation regime'

Y Combinator CEO Garry Tan has taken a contrarian stance on the controversy surrounding AI distillation, arguing that regulators should do nothing to restrict the practice and that American open-weight labs should be free to distill frontier models.

Distillation is a technique where a model maker extensively prompts another model to learn how it works and reasons. It is commonly used legitimately, but AI company Anthropic has alleged that Chinese labs are engaging in "illicit distillation attacks" — hiding their identities to distill without permission, relying on fraud and stolen credentials. Anthropic CEO Dario Amodei has publicly called on U.S. regulators to crack down on the practice.

Tan disagrees. Speaking to CNBC earlier this week, he said, "I would do nothing. We could argue that there should be an American distillation regime." In an interview with TechCrunch, he elaborated that this means smaller American open-weight labs should be able to use similar training techniques on American frontier labs, giving the U.S. a more robust set of homegrown open-weight options.

Tan is not advocating for American labs to use stolen credentials. He argued it is an overreach for frontier labs to dictate what customers can do with information their models share, and he noted that proprietary AI labs themselves did not ask permission when they ingested much of the internet's knowledge — including copyrighted material — to train their models.

"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," Tan told TechCrunch.

Tan envisions a balance between open-weight and frontier labs. "They are at the frontier and driving it forward. We want that to be fundable, and be a great business model ongoing," he told CNBC. "The nightmare scenario, the doomer scenario for AI is that there's just one company. 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."

Why This Matters

Tan's position directly opposes the stance of frontier labs like Anthropic, which have been lobbying for government action on distillation. How the Biden administration or future administrations respond could shape the power balance between open-weight and proprietary AI.
If regulators take Tan's advice, smaller open-weight labs could thrive, offering alternatives to Chinese open-weight models, but frontier labs may see their business models threatened if their models are extensively distilled without compensation.
The debate also raises legal and ethical questions about the terms of API usage and intellectual property in AI training.

Background

Tan's intervention comes amid growing tension between major proprietary AI labs and the open-weight AI community. Anthropic, a leading frontier AI company, this week released its second report alleging Chinese labs are engaging in illicit distillation attacks. The reports have spurred calls for government action, dividing even Silicon Valley.

Tan chairs Y Combinator, one of the world's most influential startup accelerators, backing hundreds of early-stage companies, many of them in AI. His perspective reflects the interests of many startup founders and open-source advocates who worry that over-regulation could solidify the dominance of large, well-capitalized AI labs.

Key Perspectives

YC & Open-Weight Advocates: Garry Tan leads the charge for a deregulatory approach, viewing distillation as a legitimate and healthy process that should be openly available. He argues it prevents a monopoly on frontier AI power. Anthropic & Frontier Labs: CEO Dario Amodei and Anthropic have identified illicit distillation as a threat, arguing that Chinese labs are using fraud and stolen credentials to bypass terms of service, and should be stopped by regulators. Critics/Skeptics: Some may argue Tan's position underplays the security and ethical concerns of allowing unchecked distillation, particularly given Anthropic's evidence of fraud. Others could also note the hypocrisy in asking regulators to keep hands off when frontier labs themselves want to maintain control over their models.

What to Watch

Whether regulatory bodies in Washington begin to take visible action — or signal they will not — in response to Anthropic's reports and Tan's counterargument.
The outcome of Anthropic's copyright settlement as an indicator of precedent for training on data without permission.
The reaction from other AI labs, including Google and Microsoft-backed OpenAI, to Tan's stance.

ZOTPAPER