When it comes to Chinese AI labs utilizing distillation to learn from frontier models, Y Combinator CEO Garry Tan hopes regulators will not interfere. Instead, he believes U.S. AI labs should adopt similar strategies.
"I would do nothing," Tan told CNBC earlier this week. "We could argue that there should be an American distillation regime." He explained to TechCrunch that this involves allowing smaller, open-weight U.S. labs to train on American frontier labs, thereby creating a more robust set of open-weight models that differ from Chinese ones.
Distillation involves a model maker extensively prompting another model to understand its internal processes and reasoning. This is a common and legitimate method used by AI companies to train new models.
This week, Anthropic released a report accusing Chinese labs of conducting "illicit distillation attacks" using stolen credentials and fraud to hide their identities. Anthropic CEO Dario Amodei had previously asked U.S. regulators to crack down on this practice. However, Tan disagrees with the commander of Silicon Valley's top startup accelerator.
It is important to clarify that Tan is not suggesting American labs use stolen credentials. He wants them to access data legitimately. His argument has two parts: he feels it is an overreach for AI labs to control what customers do with shared information. Additionally, he notes that frontier labs did not ask for permission to use vast amounts of copyrighted human knowledge during their own training.
"Controlling what users and customers do with API calls to closed weight models feels constraining," Tan told TechCrunch. He believes the government has a role in normalizing the idea that intelligence trained on public data should be a public good rather than restricted by terms of service.
Tan wants to find 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. He envisions open-weight models providing people with freedom and access.
Tan views the ultimate negative scenario as frontier AI power falling into the hands of a single proprietary provider. "The nightmare scenario... 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," he said.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.