Y Combinator CEO Garry Tan has expressed support for American AI labs using distillation methods to learn from leading AI models, a practice currently under scrutiny due to allegations of illicit use by some Chinese labs. Distillation involves extensively querying a model to understand its reasoning and behavior, a common technique for training new AI models.
Tan argues that US regulators should avoid restricting this practice and instead consider establishing a framework that allows smaller American open-weight AI labs to distill frontier models openly and legitimately. This approach, he believes, would help build a diverse and robust AI ecosystem in the US, providing alternatives to Chinese AI offerings.
His stance contrasts with recent calls from AI company Anthropic, which has accused Chinese labs of unauthorized distillation and urged regulatory intervention. Tan emphasizes that he does not support unethical behavior such as using stolen credentials but advocates for lawful and transparent access.
He further points out that proprietary AI labs themselves did not seek permission when training on vast amounts of publicly available data, including copyrighted material. Tan suggests that restricting how users interact with AI models through APIs could be overly limiting and that access to AI intelligence derived from public data should be treated more like a public good.
Tan envisions a balanced AI landscape where frontier labs continue to innovate and receive funding, while open-weight models provide broader access and freedom. He warns against a scenario where a single dominant AI company controls the technology, which could stifle competition and innovation.
By promoting open distillation practices, Tan hopes to foster a competitive and accessible AI environment in the US that counters monopolization and supports ongoing advancement.