As language models are tolerated that is only fair in my opinion. If they are allowed to be trained with GPL licensed, copyrighted and so on data then copyright should vanish also for everyone else. The existence of LLMs with the current laws is just unfair.