Summary & Key Takeaways ​

• The article explores a proposal for US AI policy regarding data collection and model distillation.
• Ben Thompson suggests explicit fair use for training data and banning TOS that forbid distillation for US companies.
• This policy aims to help US open models compete more effectively with Chinese counterparts.
• Alibaba's release of Qwen 3.8 Max as open weights is noted as a potential response to policy shifts or national directives.
• The discussion highlights the complex interplay of intellectual property, national strategy, and open-source principles in AI.

Our Commentary ​

This is a bombshell. Ben Thompson's proposal is provocative, but I can't deny its logic. The hypocrisy of labs training on unlicensed data while trying to outlaw distillation is glaring. And the idea of a US law mandating open distillation? That's a paradigm shift. It feels like the US is finally waking up to the strategic implications of open-source AI, especially with China's moves. This could fundamentally reshape the AI landscape.