Large Language Models (LLMs) and AI companies routinely use massive amounts of data for training, much of which is likely to contain copyrighted material.