I really doubt that technical arguments are going to push anybody one way or another on LLMs and copyright. At some point Congress is going to update the letter of the law to fit the spirit, and the spirit is no normal person will ever view a chatbot as a copy of its inputs. If it can reproduce its inputs verbatim, all that will be required of a vendor is to be demonstrating they're continually working to try to prevent that, e.g. how youtube whack-a-moles all manner of copyright violation