H.M. Wolfe’s Daggermouth was a self-publishing success story before it became the center of online scrutiny.
The dystopian romance topped Amazon charts, landed on the USA Today bestseller list, and attracted a seven-figure deal from Simon and Schuster. Then The Atlantic flagged it for substantial AI-written content. Wolfe vehemently denied it while her publisher stood beside her. The book became an instant #1 on The New York Times bestseller list.
But the damage to the conversation around Black authorship was already done, because Daggermouth was not the only case, and Wolfe was not the only Black author in the crosshairs.
Hachette pulled Mia Ballard’s horror novel Shy Girl after similar allegations. Jerry Falade lost a more than $2 million deal with Minotaur for his crime novel Call Me, I’ll Hide the Body, as reported by The Wall Street Journal. No concrete proof of AI use was established in any of these cases. The detection tool used against Daggermouth was an AI tool called Pangram, which produced a 60% score for suspected AI use.
Falade spoke publicly about what he saw as a racial double standard, telling The Guardian that all three Black authors who received major publishing deals that year had them canceled or disrupted over AI suspicions, while white writers who openly discussed using AI in their work faced no comparable consequences. The pattern, he argued, reflected a persistent assumption in publishing that exceptional work by Black writers must have had outside help. The double standard is part of a much larger story.
The Books Are Being Destroyed
While Black authors are being accused of using AI, AI companies are actively consuming Black literature at industrial scale. According to a July 2026 investigation first reported by 404 Media, AI companies use intermediaries to purchase books at industrial scale while keeping their identities anonymous, then destroy the physical copies after scanning. The process uses an industrial hydraulic cutting machine to slice the spine off the book before a high-speed scanner digitizes the pages. The originals are recycled, no author is notified, and no permission is ever sought.
This is Project Panama, Anthropic’s documented book digitization initiative, revealed through more than 4,000 pages of unsealed court documents reviewed by the Washington Post. Anthropic spent tens of millions of dollars secretly buying, scanning, and destroying millions of physical books to train its Claude chatbot. A federal judge ruled in June 2025 in Bartz v. Anthropic that digitizing legally purchased books constituted transformative fair use.
Where Anthropic crossed a legal line was elsewhere. Anthropic co-founder Ben Mann personally downloaded millions of pirated books from LibGen and Books3 in 2021, according to Judge William Alsup’s June 2025 order. The court found Anthropic pirated more than seven million copies of books in total. The $1.5 billion settlement that followed was significant in number and nominal in practice once dispersed across every author involved.
Anthropic isn’t alone in this. ISBNdb, which describes itself as the world’s largest book database, now brokers high-volume acquisitions for AI clients and markets pre-2022 books as premium training data — clean, human-authored, and untouched by the AI-generated content that has already saturated much of the internet. Books published before generative AI went mainstream are now a commodity, according to the 404 Media investigation. The authors who wrote them were never part of that conversation.
What Happens When AI Ingests a Book
When a book enters an AI training pipeline, the author’s voice does not survive the process intact. Jean Austin, an AI expert who provides commentary on federal AI regulations and regularly educates HBCU students about AI’s benefits and limitations, explained what actually happens. The book is broken into tokens — fragments of words rather than whole words — and the original document as a coherent object does not survive.
“The book itself is not stored anywhere in the model,” Austin said. “What’s captured instead are statistical regularities: how this author tends to structure a sentence, which words cluster together, what rhythm shows up across long passages.”
The author’s style is then blended with thousands of others, absorbed into patterns that can be mimicked but never replicated with any real authenticity.
IP attorney and author of over 20 books, Kenneth Eade, told EBONY that “An AI model cannot even ‘read’ and has no concept of story — only patterns.”
Gift Kenneth, a machine learning engineer and health-tech founder building for African women across the continent and diaspora, described the result plainly.
“The Black experience is unique to each person, depending on the social, economic, and personal factors that shape how they navigate life and interact with the world,” she said. “AI doesn’t have this perspective.”
The ingestion of Black literature is happening within a framework that was already skewed long before AI arrived. In 2024, career website Zippia reported that only roughly 6% of authors published annually in the US are Black, even as Black Americans make up more than 13% of the population. The 2020 #PublishingPaidMe movement put numbers to what Black authors had long described. Jesmyn Ward, already a National Book Award winner, had to negotiate hard for a $100,000 advance. These were the conditions before AI began consuming the books Black authors fought to get published.
Now those same books are being fed into systems designed to help anyone write anything — including Black stories — without the lived experience that makes those stories matter. Kenneth raised the question at the center of it all.
“Maybe the case isn’t that there isn’t enough data,” she said. “The case is who gets to choose what data is relevant and what isn’t.”
Black authors have spent their careers fighting for shelf space in a predominantly white publishing industry. AI is not solving that problem. It is inheriting and scaling it.
How to Spot an AI Book
Readers can take practical steps to spot AI-generated books, though none are foolproof. Eade recommends starting with the US Copyright Office since AI-only books cannot be registered, and mass producers typically don’t bother filing. If the author has no prior books, no interviews, and no verifiable public presence, that warrants scrutiny. Authors producing multiple full-length titles per month are not real people. In nonfiction, pick three cited sources and look them up. Fabricated citations remain one of the most reliable tells. Books with titles or covers too similar to existing bestsellers are another red flag.
At the prose level, Austin pointed to structure that feels too polished, vague emotional language, and overuse of em dashes where a comma would do. Real writing tends to include oddly specific detail, like the exact smell of a waiting room or the precise weight of a particular moment. AI stays general because specificity is hard to manufacture. Austin also recommended getting AI prompt-certified through courses offered by companies like Google, arguing that understanding how AI frameworks are built makes their patterns easier to recognize.
But Kenneth raised a valid point. No fully reliable signals exist, and the tools used to detect AI writing are themselves imperfect. People who learned English as a second language often write in formal, polished sentences that detection tools flag incorrectly. As AI systems improve and the lines between human and machine writing continue to blur, Kenneth warned that detection tools will struggle to keep up and will continue producing false positives.
What the Future Could Hold
Austin’s projection for the next five years is direct. The publishing landscape will be controlled by five to six companies, he said, and the marginalized voice will fade. He called on Black authors to copyright their books and trademark their brands immediately, to own intellectual property under corporate structures, and to build Black and Brown digital and print library repositories that control the narrative independent of the major publishing apparatus.
Eade framed the stakes in cultural terms. “It is important that physical libraries of actual real books be kept by the guardians of the culture,” he said, “or it will be lost by assimilation into a non-human culture that will spread a ‘machine wash’ over it.”