Language models should be able to process and learn from copyrighted text as it's essential for linguistic development. Restricting this could stifle innovation and free speech.