Y Combinator's Garry Tan wants US open-weight AI labs to 'distill' frontier models, too

Y Combinator CEO Garry Tan is calling for smaller, American open-weight AI labs to use the same training techniques on American frontier AI labs, giving the US a more robust set of open-weight options. This approach is a response to Chinese AI labs using distillation techniques to extract knowledge from frontier model makers, a practice that Anthropic CEO Dario Amodei has alleged is engaging in illicit distillation attacks. Tan argues that this approach would allow American AI labs to access the same training data without the need for stolen credentials. He also believes that controlling what users and customers do with the information their models share with them is an overreach, and that access to intelligence trained on broad public access data should be a public good. Tan wants to see a balance between open-weight AI labs and frontier labs, with the goal of making frontier AI fundable and a great business model ongoing. In contrast, a single powerful, proprietary provider of frontier AI could lead to a nightmare scenario where one company has the best access to capital and AI researchers, and runs away with the technology. Tan's proposal is part of a broader effort to promote open and accessible AI development in the US.