Jensen Huang Defends AI Distillation as Learning, Not Stealing

Nvidia CEO Jensen Huang recently defended AI model distillation and open-source AI in an interview with Axios, addressing ongoing industry controversies. He emphasized that distillation is essentially learning and the foundation of intelligence, making it wrong to view AI absorbing information from other models and knowledge sources as "stealing." The current conclusion is that Nvidia officially supports mutual learning among models as an inevitable trend in technological development, provided that privacy is not violated.

Confirmed

Huang explicitly stated that distillation is learning, just as natural as humans acquiring knowledge from books, teachers, and peers. He noted that early models learned from human knowledge on the internet, but as the internet evolves, 99% of its content might be generated by AI in the future. Therefore, AI models learning from and distilling each other will become the norm, which is also a core reason he values the importance of open-source models. Furthermore, he argued that making AI smarter also makes it safer. While advocating for broad learning, Huang drew a clear底线, stating that AI must not infringe on privacy during its learning process.

Why it matters

Recently, controversies over startups or open-source models "distilling" leading closed-source models have sparked debates regarding intellectual property and commercial competition. As the dominant force in AI computing, Nvidia's CEO publicly legitimizing "distillation" and defining it as the "foundation of intelligence" provides strong endorsement for the open-source community. It also signals that in Nvidia's vision, knowledge transfer and sharing among models are essential paths to driving AI into a safer and more intelligent era.

2026-07-26 ~ 2026-07-27 · 5 related posts

Full story(8 episodes)→

Primary sources