Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too

Live updates

1. 57m ago

   Garry Tan opposes regulation of AI model distillation

   Y Combinator CEO Garry Tan argues that regulators should not intervene in AI model distillation. Tan believes US open-weight AI labs must use this practice to stay competitive and suggests the US government avoid stopping it. This position contrasts with warnings from the NSA and accusations from AI labs like Anthropic, which claim Chinese firms such as Alibaba, DeepSeek, and Moonshot AI use distillation to copy US frontier models. China denies these allegations of industrial-scale AI theft.

   Why it matters

   Distillation involves extensively prompting one model to train another. This technique allows smaller models to mimic the performance of larger, more expensive frontier models.

   What is confirmed

   • Garry Tan is the CEO of Y Combinator.
   • Anthropic claims Moonshot AI, DeepSeek, and Alibaba conducted distillation campaigns.
   • The NSA has issued warnings regarding AI distillation practices.

   Still unconfirmed

   • Moonshot and DeepSeek secretly diverted user requests to Claude.
   • Chinese firms are aggressively copying US frontier models through distillation.

   What to watch next

   • US government regulatory decisions on AI distillation
   • Official responses from the NSA regarding Tan's position
   Sources used for this update (2)

   1. theaiinsider.tech — Y Combinator’s Garry Tan Argues Against Regulating AI Model Distillation
   2. cryptobriefing.com — Sam Altman reflects on governance lessons from Y Combinator for OpenAI
   confidence 100%
2. 1d ago

   US Accuses Chinese AI Firms of Copying Frontier Models via Distillation

   The US government and AI labs including Anthropic accuse Chinese firms of aggressively copying US frontier models through distillation. Anthropic claims Moonshot AI, DeepSeek, and Alibaba conducted distillation campaigns, alleging Moonshot and DeepSeek secretly diverted user requests to Claude. The NSA has issued warnings regarding these practices. Amid these accusations, Y Combinator's Garry Tan argues that US open-weight AI labs should also utilize distillation to remain competitive, suggesting the US do nothing to stop the practice. China has denied the claims of industrial-scale AI theft.

   Why it matters

   Distillation allows smaller models to mimic the performance of larger, more expensive frontier models by training on their outputs. This conflict highlights the tension between protecting intellectual property and the global race for AI efficiency. The dispute escalates as the US and China compete for AI influence in regions like Africa.

   What is confirmed

   • The US government accuses Chinese AI firms of malicious copying of AI technology.
   • Anthropic identifies Alibaba, Moonshot AI, and DeepSeek as participants in distillation campaigns.
   • The NSA warns that China-based AI companies are distilling US frontier AI models.
   • Six Chinese AI firms are accused of aggressively copying US frontier models.

   Still unconfirmed

   • Moonshot AI and DeepSeek secretly diverted user requests to Claude.
   • Chinese AI labs used millions of Claude exchanges to train their models.
   • Garry Tan wants US open-weight AI labs to distill frontier models.

   What to watch next

   • Legal action or sanctions resulting from the claims of intellectual property theft.
   Sources used for this update (13)

   1. Reuters — US accuses Chinese AI firms of 'malicious' copying of AI technology
   2. arstechnica.com — Six Chinese AI firms accused of aggressively copying US frontier models
   3. TechCrunch — Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek
   4. Bloomberg.com — Moonshot Secretly Routed User Requests Through Claude, Anthropic Says
   5. WSJ — How Chinese AI Firms Tried to Clone U.S. AI Models
   6. CNBC — Chinese AI labs secretly used millions of Claude exchanges to train their models, Anthropic says
   7. South China Morning Post — Anthropic claims Moonshot, DeepSeek secretly diverted user requests to Claude
   8. WSJ — The Morning Risk Report: How Chinese AI Firms Tried to Clone U.S. AI Models
   9. TechCrunch — Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too
   10. CNBC — Y Combinator’s Garry Tan says 'do nothing' about distillation as AI giants accuse China of copying their tech
   11. nsa.gov — NSA and Others Warn China-Based AI Companies are Distilling U.S. Frontier AI Models
   12. Business Insider Africa — China hits back at U.S. over industrial-scale AI theft claims as Washington beats Beijing to Africa with $1 billion AI deal
   confidence 90%