nytimes.com via Reddit

Future of Truth book cited for AI-fabricated quotes

hallucinations ai ethics ai-hallucinations publishing ai-ethics

Key insights

  • The Future of Truth, a book about AI disinformation, contained AI-fabricated quotes attributed to real people that passed editorial review.
  • EY Canada separately retracted a client report days earlier after 60% of its citations were found to be AI hallucinations.
  • Both failures share the same mechanism: AI tools inserted at the research stage produced fabrications that downstream human review did not catch.

Why this matters

Two high-profile AI hallucination failures in professional publishing and consulting within days of each other signal that the failure mode is systemic, not anecdotal, and is now producing concrete reputational and legal exposure for organizations shipping AI-assisted knowledge products. For technical leaders, the specific failure point is workflow architecture: AI tools are being used at the research and transcription stage without a verification layer capable of catching plausible, confident fabrications before they reach reviewers. The combination of a defamation-adjacent publishing case and a 60% citation hallucination rate in a client-facing consulting report will accelerate enterprise procurement requirements for AI-output auditing across both sectors.

Summary

A book titled 'The Future of Truth', written specifically to address AI-generated disinformation, was found to contain multiple quotes attributed to real people that were never actually said. The author used AI tools to locate and transcribe source quotes; those tools hallucinated statements, attributed them to real individuals, and the fabrications cleared editorial review. This surfaces days after EY Canada retracted a client report where 60% of citations were hallucinated, a second high-profile case of AI research workflows producing confident fabrications that human reviewers did not catch. Essentially: (the book's author, EY Canada) both trusted AI-generated source material without independent verification at the output stage. - The book's subject matter sharpens the failure: it was written to help readers resist AI-generated falsehoods in the information environment. - EY Canada's 60% hallucination rate in a professional client deliverable shows this is a cross-industry pattern, not an isolated publishing incident. Two high-profile failures in the same week point to a systemic verification gap in AI-assisted research workflows, not isolated operator error.

Potential risks and opportunities

Risks

  • Individuals whose quotes were fabricated could pursue defamation claims against the author and publisher if the attributed statements are materially false or damaging to their reputation
  • Publishers that have adopted AI-assisted research workflows face accelerated scrutiny from authors, insurers, and fact-checkers, with liability exposure if fabricated content remains uncorrected in printed and digital editions already shipped
  • EY Canada and peer consulting firms (Deloitte, KPMG, PwC) face client demands for explicit AI-use disclosure on all research deliverables within the next 30 to 60 days, given the dual high-profile failures in the same news cycle

Opportunities

  • Citation and fact-verification startups (Consensus, Semantic Scholar, Citationsy) have a concrete enterprise sales case for integration into publishing and professional services research workflows following both incidents
  • Publishers and consulting firms that publicly deploy and document AI verification protocols in the next 60 days can differentiate on trust at a moment when direct competitors are managing active retractions
  • AI governance and compliance vendors serving legal and professional services can price AI-output auditing services upward, with EY Canada's case now serving as a named reference point for risk quantification in sales cycles

What we don't know yet

  • Which specific AI tool the author used for quote sourcing, and whether the tool's developer has been notified or named in the reporting
  • Whether the publisher plans a correction, retraction, or revised edition, and whether printed copies already in distribution will be addressed
  • Identity of the real individuals whose quotes were fabricated, and whether any have disputed the attributions or indicated intent to pursue claims as of May 19, 2026

Originally reported by nytimes.com

Read the original article →

Original headline: NYT: Book on Truth in the Age of A.I. Contains Quotes Made Up by A.I.