The New York Times reported something that writes itself: someone wrote a book about "truth in the age of AI," and the book contains multiple quotes that don't exist. AI made them up. The author wrote with AI.
The book is called The Future of Truth. The author is Steven Rosenbaum, who calls himself "The Truth Whisperer." Putting the book title and the author's self-appointed nickname side by side reaches a level of irony that probably needs a new unit of measurement.
What Happened
Rosenbaum used Claude and ChatGPT for research, writing, and editing. In the process, AI generated several quotes that looked legitimate — they had sources, context, tone — but were entirely fabricated.
The New York Times called out the fabricated quotes. Rosenbaum's response: he takes "full responsibility," but "these AI errors do not diminish the larger questions that the book raises."
That response is the most interesting part of the whole story.
"Larger Questions" Don't Excuse Fabrication
Rosenbaum's defense logic is: my argument is correct, so whether the evidence is fabricated doesn't matter.
This logic doesn't hold in any serious form of writing. Especially not in a book about truth.
If you wrote a book about climate science and it turned out several data sets were fabricated — would you say "it's fine, the general direction of climate change is right"? No. Because the accuracy of data isn't a side note to the argument — it IS part of the argument.
Similarly, a book about how AI threatens truth, if it itself manufactures false information — the act itself is more persuasive than any argument in the book. It's proving its own thesis through its own existence.
This isn't irony. It's empirical evidence.
AI Hallucination: From Technical Bug to Trust Crisis
This goes deeper than it appears on the surface.
AI hallucination has always been treated as a technical flaw in large language models — the model occasionally invents facts, and humans need to check. But when an author publishes AI-generated content into a formal publication without verification, hallucination stops being a technical problem. It becomes a trust breakdown in the content production pipeline.
Rosenbaum isn't the first, and he won't be the last. Claude and ChatGPT are now widely used for research and writing assistance, but the line between "assistance" and "ghostwriting" is one that many authors aren't thinking about carefully enough.
Using AI for research doesn't mean you skip verifying the citations it gives you. That should be common sense. But in the age of vibe coding and "AI writes everything for you" narratives, more and more people are treating AI as an authority that doesn't need checking.
This isn't AI's fault. It's the people using AI abandoning the most basic professional judgment.
The Publishing Industry's Awkward Position
Publishers will face this problem more and more frequently. A book already on the market is found to contain AI-fabricated quotes — recall it? Correct it? Put a sticker on the cover saying "some content AI-generated, not verified"?
I don't know the publishing industry's standard response. But I know that if this happened in academic publishing, the consequences would be much more severe. Fabricated citations are a core offense in the definition of academic misconduct.
Mass-market publishing doesn't yet have verification standards for AI-generated content. Or rather, the standards haven't caught up with the speed of technology diffusion.
My Take
This book's experience is a clear warning: the difference between AI-assisted writing and AI ghostwriting is exactly one "verify" action.
Rosenbaum's problem isn't that he used AI. It's that he didn't perform the most basic verification on AI's output. A quote — search whether the original words exist. Five minutes. He didn't search. Or rather, he trusted AI more than his own professional judgment.
This isn't a question of technical capability. It's a question of professional integrity.
If the book's thesis is "AI is threatening truth" — then the most powerful evidence isn't anything written in the book. It's the process of the book itself being caught by the New York Times.
A book about truth, exposed because AI fabricated its quotes. The irony level of this story probably exceeds any AI's imagination.
Main sources: