Bottom line: Those who use unlicensed data for their own training should not be able to prohibit others from distilling their own models—a regulatory gap with competitive implications between US and Chinese AI labs.
A proposal by Ben Thompson could establish training data usage explicitly as fair use in the US and prohibit distillation bans in terms of service. The background is competition with Chinese open-source models like Alibaba’s Qwen 3.8 Max.
Tech analyst Ben Thompson has outlined a regulatory proposal that addresses an inconsistency in current AI practice: large US labs train their models on data whose use is often not explicitly authorized, yet simultaneously prohibit distillation of their trained models via terms of service. Distillation—merely querying APIs for knowledge extraction—is technically difficult to prevent.
Thompson’s proposal comprises two pillars: first, the US should legally clarify that the collection of training data falls under fair use. Second, terms of service for US companies should no longer be permitted to prohibit distillation. This would create a new copyright policy that indemnifies labs but ensures that learned techniques advance innovation across the industry.
A trigger for this debate could be a strategic shift: Alibaba recently released Qwen 3.8 Max as open weights—a change of course from the decision in May not to publish the previous version Qwen 3.7 Max. Thompson suspects a connection to a speech by Xi Jinping, who said: “We should seize this rare, historic opportunity to promote open source, openness, collaboration and exchange.” This positions China as an advocate for open models, while US labs attempt to enforce more restrictive licenses.
For Chief Data Officers, this means: regulatory questions about data use, distillation, and licensing become competitive questions. An asymmetric regulation—where US labs prohibit distillation while Chinese providers are openly accessible—could in the long term alter the technological advantage.
Source: simonwillison.net · Published 20 July 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrasing and classification via Lumi News Pipeline v1.7.3.