Rosenbaum's AI-Truth Book Ships With Fabricated Quotes

A book called "The Future of Truth: How AI Reshapes Reality," published to explore how artificial intelligence distorts what we believe, turned out to contain multiple quotes that AI had fabricated or misattributed. The author, Steven Rosenbaum, acknowledged the errors on May 19 after an inquiry from The New York Times, days after the book was published by Matt Holt Books on May 12, 2026.

The specifics matter. According to Marketplace, among those whose words were mangled was Meredith Broussard, a journalist and NYU professor. A quote she genuinely said during a 2023 interview on Marketplace Tech ended up cited in Rosenbaum's book as having appeared in her book "Artificial Unintelligence." The words were hers; the sourcing was not. Kara Swisher described one quote attributed to her as making her "sound like I have a stick up my butt." Lisa Feldman Barrett said a quote attributed to her misrepresented her views on the nature of emotions, social signals, and truth.

Rosenbaum has said he used AI chatbots as research tools during the book's development, indicating in his notes what information came from AI and sending those claims to a fact-checker affiliated with the publisher. That process did not catch the hallucinations. He is now working with editors to review and correct affected passages, with future editions to reflect the changes.

The honest caveat is that we do not yet know how many other quotes in the book remain undetected, nor exactly how the fact-checking process broke down. What the reporting does not give you is a clear account of what specific verification steps were applied, or whether corrections will reach readers who already hold the current edition.

The practical lesson is something practitioners in any research-heavy field should sit with: AI hallucinations do not announce themselves. The model does not flag that a citation is invented; it presents fabricated sourcing with the same confidence as accurate sourcing. A book specifically about AI's dangers was not immune. Publishers now have a concrete, high-profile example to justify building better AI-sourcing disclosure standards, and anyone using AI as a research shortcut in journalism, academia, or publishing has reason to pressure-test their verification pipelines more carefully.