Garry Tan Urges 'American Distillation Regime' to Protect Open AI
The Y Combinator CEO argues that access to frontier model intelligence should be a public good, resisting government crackdowns on AI distillation.
Y Combinator CEO Garry Tan is urging the U.S. government to resist crackdowns on AI distillation, arguing that the practice is essential for a healthy open-weight AI ecosystem. Tan suggests that instead of restrictive regulations, the U.S. should embrace an "American distillation regime" to ensure domestic labs can compete with proprietary giants.
Distillation is the process of using the outputs of a massive "frontier" model to train a smaller, more efficient one. The debate intensified following a September 2026 report from Anthropic, which alleged that Chinese laboratories have been conducting "illicit distillation attacks" by using stolen credentials and fraud to scrape data from Claude. While Anthropic and other frontier labs have called for regulation to stop these practices to protect their intellectual property and safety guardrails, Tan argues that the U.S. should "do nothing" to stop the process domestically.
The Battle for Open Weights
The conflict highlights a growing rift between frontier labs—such as OpenAI and Anthropic—and open-weight labs that seek to democratize high-capability AI. Tan contends that intelligence trained on broad public data should be treated as a "public good" rather than a proprietary asset locked behind restrictive terms of service. He suggests that the current restrictions imposed by proprietary labs are inconsistent with how those same companies originally built their models, noting that they did not seek permission when ingesting copyrighted material for their own training.
Preventing an AI Monopoly
For Tan, the stakes involve the long-term structure of the AI industry. He warns that the "nightmare scenario" for the field is a monolithic market where a single company controls the most advanced capabilities. By normalizing or legally protecting distillation, the U.S. could accelerate the development of high-performance open-source models, preventing a scenario where a few proprietary firms hold an absolute monopoly over artificial intelligence.
The Path Forward
As the U.S. government weighs how to respond to the security concerns raised by Anthropic regarding foreign adversaries, the policy decision will determine whether distillation remains a legal gray area or becomes a protected tool for domestic innovation. It remains to be seen if the government will prioritize the intellectual property protections requested by frontier labs or the open-access framework proposed by Tan. This decision will ultimately shape whether the next generation of AI is concentrated in a few corporate hands or distributed across a diverse ecosystem of open-weight developers.