I suspect the truth lies somewhere in the middle. AI models are trained on material from the Internet, much of which is under copyright. However, it does not COPY these materials but rather draws statistical inferences from them. The models do not contain the original data. But...at the same...