Garry Tan, CEO of Y Combinator, oversaw the organization’s Fall 2026 Demo Day amidst growing industry debate over model distillation. The process has become a flashpoint as American developers and international competitors navigate the evolving AI landscape.
Distillation is a process where a smaller, more efficient “student” model is trained using the outputs of a more powerful “teacher” model to replicate its capabilities at a fraction of the original training cost. While companies like OpenAI and Anthropic explicitly prohibit the use of their model outputs to train competing systems in their terms of service, the practice remains a central topic of discussion regarding how to maintain a competitive baseline against proprietary systems.
The debate over distillation comes as U.S. intelligence agencies track foreign exploitation of American AI infrastructure. On September 8, 2026, the NSA, FBI, and CISA issued a joint advisory naming six Chinese entities—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—for conducting “industrial-scale” distillation campaigns against American frontier models.
Recent data from proprietary labs suggests these campaigns are high-volume and persistent. Anthropic reported on September 11, 2026, that it blocked 151 million exchange attempts from Alibaba-linked operators between May and July 2026, characterizing the activity as an illicit distillation attack. Similarly, OpenAI has reportedly concluded that the architecture of DeepSeek’s V3 and R1 models was significantly influenced by outputs from GPT-4 and GPT-4o.
The practice of distillation has drawn criticism from the leadership of several major proprietary labs. Anthropic CEO Dario Amodei has called for U.S. regulators to implement strict legal barriers against distillation, viewing the practice as a bypass of the massive research and development investments required to build frontier models.
The debate underscores a deepening rift in Silicon Valley between those who prioritize open-source resilience and those who argue that uncontrolled model proliferation poses commercial and security risks. The challenge of regulating distillation remains complex, particularly as international actors utilize the technique to achieve parity with U.S. technology.
