Practical applications for AI in publishing have existed for a number of years now in areas such as statistical review and copyediting, but the recent and prominent emergence of AI authorship presents areas for both vast potential and great concern. Where are the boundaries for AI use in scholarly publishing, and who should be setting them?