Photo: Tima Miroshnichenko / Pexels
The accelerator chief argues that AI models trained on public knowledge should remain accessible, even as US agencies warn about Chinese distillation efforts
Garry Tan, president and CEO of Y Combinator, used the stage at the accelerator’s Demo Day to deliver a pointed message to regulators: leave open-weight AI alone. Speaking on September 10, Tan argued that frontier AI models, built on the bedrock of public human knowledge, should function as “a form of public good.” The timing was deliberate, arriving just days after a joint advisory from the NSA, FBI, and CISA flagged alleged large-scale model distillation efforts by Chinese companies.
Tan’s response to that advisory was essentially a shrug. He said he would “do nothing” about Chinese firms like DeepSeek, Moonshot AI, and MiniMax reportedly distilling outputs from models built by OpenAI and Anthropic. Rather than treating distillation as an existential threat, he framed it as an inevitable byproduct of building powerful AI on publicly available data.
Open weights over closed gates #
Tan urged regulators to strike a balance that preserves “freedom and access” in AI development. His argument is that clamping down on distillation, the process of training smaller models to mimic the behavior of larger ones, would ultimately hurt the broader ecosystem more than it would protect any single company’s intellectual property.
Tan’s counterargument is structural. If frontier labs want to protect their advantage, they can do it through pricing strategy rather than regulatory moats.
Demo Day paints a clear picture #
The composition of Y Combinator’s latest batch tells its own story. Of the 196 startups that presented at Demo Day, 149 were machine learning or AI ventures. That’s roughly 76% of the cohort.
Tan has backed up his rhetoric with action. He has promoted open-source AI tools through projects like GBrain and GStack, positioning himself and YC as advocates for a more accessible AI development landscape. His push for “greater aggressiveness in pursuing open-source models” signals that this isn’t just philosophical musing.
The geopolitical backdrop #
The joint NSA-FBI-CISA advisory named specific Chinese firms, turning what had been an open industry secret into an official national security concern. By dismissing what he characterized as doomsday concerns around AI advancement, Tan redirected the conversation toward practical issues: cybersecurity, infrastructure protection, and making sure the US maintains its innovation edge through openness rather than restriction.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our