Training AI models on copyrighted works is "squarely what transformative use is," EFF’s Tori Noble told Reason - these are "general purpose tools" and "exactly the kind of creation that copyright is designed to foster.” https://reason.com/2026/09/04/doj-says-barring-ai-training-on-copyrighted-material-could-severely-hamper-tech-progress/
@eff Surely If it's just a tool, it's still possible to infringe copyright with that tool.
It seems to me, what matters is the relationship sources and the result.
Whether something is transformative is assessed case by case, creation by creation. (Irrespective of tools used).
And in the case of LLMs, where 'creators' do not select/understand their sources, whether it's copyright infringement or not is just a spin of the roulette wheel. The problem is now everyone's got infinite spins