The Future of Truth is a book about how AI is reshaping truth. Published on May 12, 2026, it carries a foreword by Maria Ressa, the Nobel Peace Prize–winning journalist. Within a week of its release, the New York Times had found more than six quotes in the book whose sources were fabricated or misattributed. One was pinned to the tech journalist Kara Swisher, who said flatly that she had never said it. The sentence had been generated by AI.
The author, Steven Rosenbaum, explained that it was an accident, not deliberate fabrication. Early in the writing process, he said, he used AI as a research tool, flagged the AI-sourced material separately, and sent it to the publisher's fact-checking contact. The fake quotes made it into the book anyway. A book meant to warn how AI threatens truth collapsed in precisely the way it warned about.
It's easy to file this under the familiar alarm: AI makes things up. But peel back one layer and a different structure comes into view. AI didn't create a new problem. It exposed a gap that publishing had left empty for a long time.
A System Where No One Owns Verification
In nonfiction publishing, fact-checking is not a contractual obligation. According to New York Magazine, which reported on the incident, hiring an outside fact-checker runs from $7,000 to more than $10,000 per book. Because of that cost, the step has long been skipped. In the same report, a senior nonfiction editor at a major publisher recalled finding errors in a book and raising the fact-checking question—only to be told that publishers bear no responsibility for fact-checking or for hiring outside verification, and that doing so would shift liability onto the publisher.
Authors are unlikely to spend heavily on verification out of their own pockets, either. One literary agent put it simply: there is no established system. The issue comes up in every publishing deal, and no one has an answer. Few publishers have specific contract language about AI. One editor said she sends authors guidance not to use AI in certain ways, but admitted it carries no legal force—once a manuscript enters editing, it ultimately comes down to trust.
Responsibility for verification is anchored nowhere. Publishers say they have no obligation, authors can't afford it, and agents say there's no system. In ordinary times, the gap never showed. At the speed humans write, errors crept in rarely, and that much was tolerated. When AI cranked up the speed of production, the vacant responsibility turned into an accident.
The Publish-First Era Was Foretold
This structure was diagnosed before AI ever arrived. In Too Big to Know, David Weinberger argued that when the internet eliminated the cost of publishing, it changed how knowledge itself works. A publish-first, verify-later strategy took hold, he observed, producing vast piles of data released before verification and accessible to anyone. The role facts had played as the foundation of knowledge shifted accordingly.
What Weinberger saw was that when the cost of publishing approaches zero, fact-checking is the first thing squeezed out. In the print era, the sheer cost of printing and distribution was itself a barrier to publication. Because every copy cost money, publishers had to choose what to print, and review was built into that choosing. The internet erased the cost. Once publishing became free, the filtering step vanished along with it.
AI went one step further. If the internet made publishing cheap, AI made production cheap. Now that one person can produce a book-length draft in a matter of days, fact-checking alone remains expensive and slow. The work requires a human to check claims against original sources one by one, which makes it hard to automate, and the thousands-of-dollars-per-book economics haven't changed. Production has accelerated severalfold; verification can't keep pace. The wider the gap between those two speeds, the greater the odds that an unchecked sentence hardens into a book.
Rosenbaum's book is the accident that gap produced. AI generated sentences quickly, verification was fixed nowhere, and the result hardened into a book and went to market. If a human had been writing line by line, six fake quotes would have had a hard time accumulating in a single volume. When production outruns verification, errors stop being rare slips and become what the structure reliably produces.
Solo Creators Have No One to Hand Verification To
Here, the position of solo entrepreneurs and independent creators is even starker than a publisher's. A publishing house at least has layers: editors, outside fact-checkers, legal review. For a solo creator, all of those layers converge on one person. There is no one to pass verification to.
Someone who self-publishes ebooks, writes their own newsletter, and drafts their own video scripts controls every stage of production alone. With AI, that work runs several times faster. Where a single piece once took days—from research through first draft—the process now compresses into hours. The problem is that verification doesn't speed up to match. When you dug up and read sources yourself, you wrote with the source in hand; when you write from AI summaries, sentences that never passed through an original source flow straight into the piece. The step publishers skipped for lack of money, solo creators are tempted to skip for lack of time and hands. The result is the same: unverified sentences go out into the world under your name.
Rosenbaum had a publisher and a distributor behind him, and the responsibility still fell entirely on him. In his statement, he accepted full responsibility for the errors. A solo creator doesn't even have anyone to share that burden with. Even if AI wrote the sentence, the person whose name is on it answers for it. Putting your name on something means vouching not for where the words came from, but for whether they're true.
So how far should verification go, and how? Sending every sentence to outside experts isn't realistic for a solo creator. The better move is to narrow the scope and concentrate.
Where to start? The spot where Rosenbaum's accident began points to the answer. The fakes came from direct quotations—words placed in Kara Swisher's mouth that she never spoke. Quotes are what AI fabricates most convincingly. Anything inside quotation marks is safest checked against the original source: did this person actually say this? Next come numbers and dates. AI presents specifics like "about 70%" or "2023" with great confidence, but that confidence and accuracy are two different things. Statistics, dollar figures, and years should be cross-checked against primary sources. Last are proper nouns. Names of people, companies, books, and events read smoothly even when AI gets them wrong, so they slip past easily. They deserve one pass to confirm they actually exist. Quotes, numbers, proper nouns: these three are where AI's errors leak out most often, and where a human can most reliably stop them before publishing.
There's a reason to focus on exactly these three. Each is a question with a single right answer, so true-or-false can actually be checked. Interpretations and opinions, by contrast, can't be settled by comparing against a source. For a solo creator who can't grow extra hands, it's rational to pour the available time into items that can genuinely be verified. Flag everything drawn from AI, and build a fixed step into your workflow: before publishing, revisit only the flagged sentences. Rosenbaum got as far as flagging the AI-sourced material. What was missing was the step that checks what's been flagged.
You Got the Speed—Now Redesign the Verification
AI handed solo creators production power that rivals a publishing house. It did not hand them the publisher's verification layers along with it. Production can be delegated; the responsibility of putting your name on something cannot.
What Rosenbaum's case leaves behind is not a lesson to stop using AI. He says he used AI only as a research tool—but he never fixed the responsibility for verification on anyone. The faster the production tools in your hands, the more important it becomes to decide up front who verifies, when, and how. The gap publishers left empty for cost reasons is one a solo creator can design back into their own workflow. If you can't place verification outside your operation, the realistic alternative is to make the final pre-publish check a fixed, standing procedure.
The era of publishing before verifying has arrived, just as Weinberger predicted. Kara Swisher learned that words she never said had been printed in a book only after the fact. If someone had checked that single sentence against its original source before publication, the book would not have collapsed. When you put your name on something you release into the world, deciding who performs that check—and when—is now part of the work itself.
