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

TechCrunch reported on September 11 that Y Combinator's Garry Tan has publicly argued that U.S. open-weight AI labs should be allowed to distill frontier models as well, pushing a technical practice to the centre of an argument about industry rules.

Tan's argument is straightforward. He contends that frontier models were themselves trained on public human knowledge, which makes access to capable AI a form of public good. If closed frontier models can benefit from that shared body of knowledge, his reasoning goes, open-weight labs should not be treated as an exception when they reach comparable capability through distillation.

Distillation generally means using a stronger model to generate training data or guidance signals for a smaller, cheaper and more easily deployed one. It is one of the most common engineering techniques in model development, and it sits at the heart of the question of whether open-weight models can keep pace with the frontier.

The dispute is really about boundaries. Some in the industry treat distillation as an efficiency tool; others worry that it erodes the returns funding expensive frontier training and therefore weakens the incentive to keep investing. By staking out a position, Tan is picking a side in that argument.

What makes his framing notable is the public-good language rather than a straightforward competitive narrative. The point is not whether one lab saves on training costs, but whether access to capability should be spread more widely. If more of the U.S. open camp adopts that framing, it will shape how model licences and usage terms are discussed.

For open-weight advocates, the statement is an endorsement from inside the U.S. startup ecosystem. Open model supporters have long had to answer two lines of criticism: safety concerns, and the charge that open weights undercut frontier labs' business models. Defining capability access as a public good sidesteps the first and engages the second directly.

The next thing to watch is how frontier labs and the open camp respond. If the public-good frame gains traction, restrictive clauses on distillation, licence design and self-imposed industry rules may come under fresh scrutiny; if nobody picks it up, it remains a single public comment.