Garry Tan, CEO of Y Combinator, publicly challenged US regulatory efforts to restrict open-weight AI models during the accelerator's Demo Day on September 10. He argued that models trained on public data should be treated as a public good, dismissing concerns regarding the distillation of AI outputs by Chinese firms.
The statement follows a joint advisory from the NSA, FBI, and CISA, which identified Chinese companies like DeepSeek, Moonshot AI, and MiniMax as entities distilling models from OpenAI and Anthropic. Tan stated he would take no action against these practices, suggesting that companies should rely on pricing strategies rather than regulatory barriers to maintain their competitive edge.
Y Combinator’s latest cohort highlights a significant focus on the sector, with 149 out of 196 startups, or approximately 76%, centered on machine learning or AI. Tan is actively promoting open-source development through initiatives such as GBrain and GStack to ensure continued accessibility in the field.
Tan advocates for maintaining US innovation through openness and infrastructure security rather than restrictive policies. He contends that limiting model distillation would negatively impact the broader technology ecosystem more than it would protect individual intellectual property rights.
