Y Combinator CEO Garry Tan has expressed support for U.S. AI labs employing distillation methods similar to those used by Chinese AI developers. Distillation involves extensively querying a model to understand its reasoning and behavior, a common practice for training new AI models. While some, including Anthropic CEO Dario Amodei, have called for regulatory action against what they describe as unauthorized distillation by Chinese labs, Tan advocates for a more open approach in the U.S.
Tan clarified that he does not endorse illicit activities such as using stolen credentials for distillation. Instead, he believes American AI labs should have legitimate access to frontier models to foster a robust ecosystem of open-weight AI options. He argues that proprietary AI developers should not overly restrict how users interact with their models, especially since these models were trained on vast amounts of publicly accessible data, often without explicit permission from content owners.
He envisions a balanced AI landscape where frontier labs continue to innovate and attract funding, while open-weight models provide broader access and freedom for users. Tan warned against the concentration of AI power in a single dominant company, which he sees as a potential risk to the industry’s diversity and progress. His stance highlights the ongoing debate over how to regulate AI development while encouraging innovation and competition.