“OpenAI's practice of crawling the entire web for training data is considered to be in a legal gray area, with arguments for it being either fair use or massive copyright theft.”