“The AI industry has largely exhausted the readily available training data from the public internet, such as the Common Crawl dataset.”