Garry Tan Urges U.S. AI Labs to Distill Frontier Models as Anthropic Alleges 2nd China Attack
Updated
Updated · TechCrunch · Sep 11
Garry Tan Urges U.S. AI Labs to Distill Frontier Models as Anthropic Alleges 2nd China Attack
2 articles · Updated · TechCrunch · Sep 11
Summary
Garry Tan said U.S. regulators should not crack down on model distillation and instead let smaller American open-weight labs distill frontier models through normal access channels.
Tan argues closed-model companies should not control how customers use model outputs, especially after proprietary labs trained on broad public data that included copyrighted material.
He drew a line at fraud, saying he is not endorsing stolen credentials or covert access even as Anthropic this week accused Chinese labs of illicit distillation in its second report.
The Y Combinator chief framed the issue as market structure: more U.S. open-weight options would counter Chinese alternatives and reduce the risk that one well-funded proprietary AI company becomes dominant.