The Irony of 'Future of Truth': Author's AI Use Undermines Book's Core Message

The Credibility Crisis at the Heart of ‘The Future of Truth’

A book investigating how artificial intelligence distorts reality has itself become a case study in AI’s impact on truth. According to Wired AI, Steve Rosenbaum’s newly published book ‘The Future of Truth’ contains over a half-dozen fabricated or misattributed quotations—a discovery The New York Times made after Wired published an excerpt in May 2026. The irony cuts deeper: the book was flagged as substantially AI-generated by multiple detection tools, forcing uncomfortable questions about whether an author can credibly critique AI’s role in undermining truth while relying heavily on the same technology.

Rosenbaum, who holds a master’s degree in “truth” from New York University, initially acknowledged using AI tools including ChatGPT, Claude, NaturalReaders, ProWritingAid, and Grammarly for “refine and polish the presentation of [his] ideas,” according to the book’s acknowledgments section. However, when Wired AI asked directly whether he had used AI to generate portions of the text, his response was equivocal: he framed AI use as “parts of the research and editorial development process” for “source discovery, brainstorming, structural feedback, and language refinement,” while insisting the “ideas, reporting, arguments, and final authorship” were his own.

Detection Tools vs. Author Claims

Wired’s investigation escalated when readers flagged the published excerpt as “blatantly AI-written,” and the publication ran it through three AI-detection services: Pangram, GPTZero, and ZeroGPT. All three returned findings suggesting the text was either likely or highly AI-generated. A subsequent analysis of the full book using Pangram—described by Wired as “the current gold standard” among detection tools—yielded a 53% AI-generation score, with an additional 9% flagged as likely AI-assisted.

Rosenbaum disputed these results, cautioning editors about false positives in AI-detection systems. He is correct that detection tools remain imperfect—they can misidentify human-written text, especially prose revised multiple times or edited by grammar-checking software. Yet the consistency of flagging across three independent services, combined with the documented quote fabrications, creates a pattern difficult to dismiss as mere algorithmic error. The publisher, BenBella Books (distributed by Simon & Schuster), did not respond to Wired’s requests for comment.

Disclosure Gap and Publishing Standards

The underlying tension reveals a gap between disclosure and accountability in AI-assisted publishing. Rosenbaum did acknowledge his use of AI tools—a step many authors skip entirely—but the nature and extent of that use remained ambiguous. When tools designed to refine language, generate brainstorms, and suggest structural revisions are applied at scale across a manuscript, the boundary between “assisted writing” and “generated writing” becomes philosophical rather than factual. The detection tools suggest the boundary has been crossed; Rosenbaum’s characterization suggests it has not.

Why This Matters

This case exposes a credibility gap that will shape how publishers, readers, and AI platforms approach transparency in the coming year. If a book explicitly about AI’s threat to truth cannot credibly document its own production methods, reader trust in AI-assisted publishing collapses—even when AI use is disclosed. For publishers developing editorial standards around generative AI, the incident suggests that acknowledgment sections and author assurances are insufficient; detection audits, third-party fact-checking, and granular disclosure of which passages received which tools may become baseline practice. For authors using AI tools ethically, the reputational risk of ambiguity—regardless of actual intent—is now visible. And for readers, the episode is a reminder that a book’s subject matter does not exempt its methods from scrutiny.