It sounds like they might be under the impression that having any AI-generated output in the code even if parts are human authored would invalidate the copyright, which isn’t true