Wired retracted a 1,450-word excerpt of Steve Rosenbaum's book The Future of Truth on Friday afternoon, after the AI-detection service Pangram scored the full manuscript as 53 percent AI-generated and an additional 9 percent likely AI-assisted. The book, published under the BenBella Books imprint and distributed in the US by Simon & Schuster, argues that artificial intelligence is warping public perception of reality. The New York Times had earlier reported that the book contained over a half-dozen made-up or misattributed quotes.
The excerpt ran earlier this month and was cleared by Wired's fact-checking team before publication. A reader email describing the piece as 'blatantly AI-written' prompted a second look. Wired ran the text through Pangram, GPTZero and ZeroGPT; each flagged it as either likely AI-generated or AI-generated with high confidence.
Rosenbaum, who holds a master's degree in 'truth' from New York University, acknowledges in the book that ChatGPT, Claude, NaturalReaders, ProWritingAid and Grammarly helped 'refine and polish the presentation' of his ideas. Pressed by Wired's head of research on how exactly those tools were used, he responded by email defending his process.
Key facts
- 01Wired retracted a 1,450-word excerpt of Steve Rosenbaum's The Future of Truth on Friday after questions about AI use.
- 02Pangram scored the full book as 53% AI-generated, with another 9% registering as likely AI-assisted.
- 03The New York Times previously reported the book contained more than half a dozen made-up or misattributed quotes.
- 04Rosenbaum's acknowledgements credit ChatGPT, Claude, NaturalReaders, ProWritingAid and Grammarly as writing aids.
- 05Hachette canceled a US novel release earlier this year after Pangram flagged it as largely AI-generated.
He also told Wired the excerpt 'was not generated by AI and then simply published as-is' and urged editors to be cautious about AI-detection tools, which can produce false positives. Wired reporter Kate Knibbs then ran the entire book through Pangram, currently considered the most reliable detector available, and got the 53 percent and 9 percent figures.
“Like many writers working today, I used AI tools during parts of the research and editorial development process for the book, including source discovery, brainstorming, structural feedback, and language refinement.”— Steve Rosenbaum, Author, The Future of Truth
Asked to address the Pangram score directly, Rosenbaum declined. 'I don't participate in that conversation,' he said. 'It's like saying, do you beat your wife. It's one of those accusations that there's no response to.' He offered instead to describe his workflow, demonstrating how he uses ChatGPT as a research tool by asking it to describe Knibbs, then reading the output aloud.
When Knibbs asked whether he would copy and paste AI output and then edit it into the manuscript, Rosenbaum answered 'probably.' Asked whether he had done so in this book specifically, he said: 'I don't remember. You're looking for a smoking gun, and there isn't one.' BenBella Books did not respond to requests for comment, and Simon & Schuster declined.
Rosenbaum repeatedly cited a MuckRack study placing AI adoption among journalists at 82 percent to argue that scrutiny of his work was selective. When Wired contacted MuckRack for a methodology breakdown, the firm said only a quarter of the writers it polled had used AI tools for writing assistance specifically — the rest of the 82 percent figure rolls in uses such as transcription and ChatGPT-based research.
Asked whether he would rather stop writing than stop using AI in his process, Rosenbaum said 'yeah.' He described AI as the best writing partner he has ever had and said writers who 'wake up in the morning wanting to have ideas' are now 'living in fear, and it's not healthy for democracy.'
The Rosenbaum episode lands in the middle of a broader publishing-industry argument about acceptable AI use. Hachette canceled the US release of a novel earlier this year after Pangram flagged it as largely AI-generated, even after the author denied using AI. The New York Times severed ties with a freelancer for AI use. On the opposite side, Fortune is actively encouraging at least one reporter to co-write stories with chatbots, Business Insider permits ChatGPT on drafts, and independent journalist Alex Heath feeds his reporting notes to an AI agent that produces his first draft.
Detection accuracy is the load-bearing question and it is genuinely unsettled. Pangram, GPTZero and ZeroGPT can disagree on the same text, and all three carry false-positive rates that publishers are still calibrating against. The Hachette and Rosenbaum cases both turn on Pangram readings the author disputes, which is exactly the kind of evidentiary question courts and contracts will eventually have to standardize.
What makes the Rosenbaum case awkward for the AI-in-publishing debate is that it collapses two separate failures into one. The Times's earlier finding of fabricated and misattributed quotes is a sourcing problem that exists with or without AI; the Pangram score is a disclosure problem about how the prose itself was produced. The book's own thesis — that AI is destabilizing the line between authentic and synthetic content — now reads as a case study in why publishers and outlets are about to invest much more heavily in provenance tooling, standardized AI-use disclosures and contractual remedies. Detection vendors like Pangram are positioned to benefit; authors who treat AI assistance as a private workflow detail are positioned to keep getting retracted.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
