Ben Thompson Proposes US Law to Counter Chinese Open-Source Models

Ben Thompson proposes a US law that makes AI training data collection fair use and prohibits terms of service that ban distillation (using one model's API to train another).

Distillation allows researchers and smaller companies to use large open-source models (like Qwen) to train smaller, more efficient models, fostering innovation and competition.

Alibaba's decision to release Qwen 3.8 Max as open weights (after not releasing Qwen 3.7 Max) may be driven by Chinese government encouragement of open source, increasing pressure on US companies to adopt similar policies.

Yes, if enacted, it would be easier for regional entities to use and distill US open-source models for local applications, reducing reliance on Chinese models and enhancing digital sovereignty.