Aaron Levie on Why Open AI Wins
Saturday, 5 September 2026 · 2 min read · Listen to the episode ↗
In this episode, Aaron Levie explores the competitive edge of open AI models, arguing that open weights drive innovation and broaden applications. He emphasizes the necessity for the U.S. to invest in open models to stimulate competition, while also addressing safety concerns. Levie predicts that as AI task costs decline, the appeal of closed models will diminish, and he highlights the potential of emerging open models like Opus 5 to enhance knowledge work efficiency.
Aaron Levie discusses the advantages of open weights in AI, asserting that they drive innovation and expand use cases. He emphasizes the need for the U.S. to invest more in open weights models to foster competition and innovation among companies. Levie argues that open models compel closed providers to innovate more rapidly, although he acknowledges the ongoing debate about the safety implications of open weights.
Levie believes that training AI models on public internet data does not cross ethical boundaries and sees no significant difference between using such data and outputs from other AI models. He predicts that a substantial investment in an open lab in the U.S. could resemble a moonshot project, although he raises concerns about national security risks associated with reliance on Chinese open weight models.
Eileen Moore adds that blocking China from AI development will not hinder its progress, as the country views AI as strategically important. She notes that the U.S. does not have an overwhelming lead in AI talent and warns that if China becomes a strong player in AI, it could harm the U.S. economy in the long run. Moore advocates for the U.S. to participate in AI infrastructure development rather than impose restrictions.
Levie highlights that the cost of AI tasks will decrease over time, making the case for closed models less compelling. He mentions that Anthropic's decision to remain closed source is likely driven by safety concerns rather than economic strategy. He also points out that Opus 5 shows significant improvements over its predecessor and is expected to be a strong competitor in knowledge work.
Levie notes that while closed labs will continue to lead in profits and advancements, the emergence of faster-follow open weights models could help them stay competitive. He expresses concerns about the current risk framework for AI being premature and emphasizes that AI has transformed product planning and execution at Box. Levie advocates for a model routing strategy to improve efficiency and cost-effectiveness in enterprises, as the proliferation of AI models can lead to analysis paralysis.
This summary was generated from the episode transcript and can contain mistakes.