Again, they haven’t made a decision on the distribution part of the training data, which is the one related to piracy. They only ruled that using copyrighted material for training LLM did not satisfy copyright infringement for the specific argument used by the authors against Meta.