Walk me through whether training on Common Crawl is legal.

The legality of training on Common Crawl relies heavily on the fair use argument. While most content within the crawl is copyrighted, the transformative nature of LLM training is the primary legal defense. The strength of this argument depends on whether the model is used for commercial or non-commercial purposes and whether the model memorizes verbatim text, which could undermine the fair use claim. Currently, there is no definitive legal precedent specifically governing LLM training.