We recently test drove ChatGPT on writing a press release for a new medical book, an assignment that, during the pen-centric era of prose, would have taken a professional medical copywriter a few hours and cost several hundred dollars. As food for “thought,” we fed the AI algorithm the Table of Contents, chapter abstracts, first and last page of every chapter, and instructions to create a press release according to a news template — the old top-down news story format with the five “W’s” in the lead — and within minutes, the program delivered a nicely formed, short, largely accurate press release that took the subject matter expert and proofreader minutes (under half an hour total) to whip into shape and submit to an online press release platform. Presumably the next one will go even more quickly, especially if it is in a similar discipline. Hundreds, maybe thousands of dollars and hours of work melt into minutes and pennies, and this is only the beginning. AI learns over time, and doesn’t forget. No looking back now! We understand that the copyright for the press release belongs to the human author who fed the request and relevant data into ChatGPT, but that is not the main issue. We think the reader needs to know the provenance of what s/he reads, needs to understand if it is machine- or human-generated — or a mixture of both. We eagerly await government-created, enforceable standards and guidelines for the use of generative language AIs in the publishing sphere, before the Tower of Babel we daily build gets too high, too cacophonous.