Black and gold cover showing a manuscript, torn publishing deal, and chain-of-custody timeline.

His $2 Million Book Deal Collapsed Over AI Doubts. Here Is What Every Author Should Learn

Jack Righteous

AI Writing · Publishing Trust · Original reporting July 2026 · Reviewed November 19, 2026

Jerry Falade’s crime novel reportedly attracted a US publishing offer exceeding $2 million before its agents withdrew the manuscript amid questions about AI use. Falade maintains that he wrote the book himself; in a later interview he acknowledged AI-assisted research, not AI-written prose. His story raises difficult questions about cultural bias, evidence and what authors should—and should not—be asked to prove.

Build the 7-File Starter RecordRead What Happened

Important: The allegation that this manuscript was AI-written is not established by the reporting cited here. Falade denies using AI for the novel’s writing or editing, but said in September he used a locally hosted model for background research. He also raised concerns about racial and cultural bias. This article does not claim the book was AI-written or treat his description as an independently adjudicated finding.

A Dream Debut Became a Publishing Crisis

A debut crime novel titled Call Me, I’ll Hide the Body appeared to have achieved the kind of breakthrough most writers can barely imagine.

According to The Guardian on July 31, 2026, the manuscript drew a competitive 14-way US auction. Minotaur, an imprint of Macmillan US, reportedly offered more than US$2 million. Substantial UK interest was also reported, with publication discussed for 2028. The figure describes a reported offer, not a publicly verified final payment.

Then the manuscript was withdrawn.

The agents’ reported explanation was unusually revealing. They said they were no longer able to authenticate how the manuscript had fully evolved “from origin to completion.” They had initially trusted Falade’s assurance that AI had not been used as a writing or editorial resource. After further questions and a July 29 meeting, the agency decided it could no longer substantiate that account and ended the relationship.

Falade denied the allegation. He argued that the suspicion reflected racial bias in publishing and pointed to other Black authors whose major deals had been disrupted by AI accusations. He also challenged the reliability of AI-detection systems and noted that the novel had passed through numerous experienced readers, editors and acquisitions teams.

What is established—and what is not

  • Established: The manuscript generated extraordinary publisher interest and a reported offer exceeding $2 million.
  • Established: The agents withdrew it after saying they could no longer authenticate its full development.
  • Established: Falade denies AI-written prose, says racial bias shaped the scrutiny and later disclosed using AI for research.
  • Not established publicly: That AI wrote the manuscript.
  • Not established publicly: That any detector conclusively proved AI-written prose or that the dispute received a definitive independent finding.

The distinction matters. This is not a clean story about software catching a dishonest writer. It is a story about trust collapsing when the creation history of a valuable manuscript became disputed.

September 16, 2026 · What the author said next

A later interview changed the story's context

In The Independent’s September 16 interview, Falade said he used a locally hosted AI model to research factual questions related to his fictional story, while maintaining that AI did not write his novel. That distinction is essential when a publishing agreement speaks vaguely about “AI use.”

He described his writing as shaped by Yorùbá oral storytelling traditions, including patterns of metaphor and rhetoric that he believed readers had wrongly interpreted as AI-generated style. He also described keeping handwritten notebooks and multiple drafts and said he had signed with a new agent at the time of the interview.

Separate his account from established findings: the September reporting records his explanations. It does not independently determine who was right about the manuscript or establish the book's later publication outcome. Publishers need clear categories for research assistance, editing and generated prose—plus consistent evidence standards that don't turn an unfamiliar cultural voice into a presumption of misconduct.

The Real Warning Is Bigger Than One Author

The easiest reaction is to argue about whether Falade used AI. The more useful question for writers is harder:

Could you explain how your own manuscript moved from its first idea to its final version?

Many authors could describe the process from memory. Far fewer could produce dated notes, early outlines, draft milestones, tracked revisions, editor correspondence, research records and clear AI-use notes.

That did not always matter. A writer could deliver a manuscript, answer editorial questions and rely on professional trust. Generative AI has complicated that relationship. Publishers, agents, clients and platforms now face questions about authorship, originality, disclosure, rights and contractual promises that were rarely asked in the same way before.

The result is uncomfortable: even a truthful author may struggle to prove a negative. A writer who did not use generative AI may still have no perfect method for proving that absence. A writer who used AI responsibly may be unable to explain where assistance ended and authorship began.

Documentation will not solve every dispute. But having no reliable record leaves the author dependent on memory, reassurance and the willingness of other people to believe them.

The Phrase Authors Need to Know: Manuscript Chain of Custody

“Chain of custody” is traditionally associated with evidence: a record showing where something came from, who handled it and how it changed. For authors, the concept can be adapted into a practical publishing habit.

Manuscript chain of custody

A manuscript chain of custody is an organized record showing how a book moved from the author’s original idea through research, outlining, drafting, AI assistance, human rewriting, editorial review and final approval.

This is not a legal certificate. It does not automatically prove copyright, guarantee publisher acceptance, prevent false accusations or settle contractual disputes. It does something more basic and immediately useful: it helps the author explain the work with evidence instead of relying only on recollection.

What a Credible Manuscript Record Can Show

A strong record does not require surveillance of every keystroke. It should preserve the milestones that reveal the human creative path.

  • Origin: dated concept notes, voice memos, journals, character sketches, framework notes or a reader promise.
  • Structure: early outlines, chapter maps, scene plans, research questions and major reorganizations.
  • Development: draft versions that show the manuscript changing over time.
  • Human decisions: notes explaining why sections were added, removed, reordered, rewritten or rejected.
  • AI involvement: the tools used, the tasks assigned to them and whether generated material entered the final work.
  • Sources: research materials, quotations, images, permissions, licences and public-domain checks.
  • Editorial history: tracked changes, editor communications, feedback rounds and author responses.
  • Approval: the final manuscript, final review date and the person who accepted responsibility for publication.

The goal is not to create a mountain of files nobody can understand. The goal is to preserve a coherent path.

Why Saving Every Prompt Is Not the Answer

Some creators hear “document your AI use” and assume they should archive every chat, prompt and generated paragraph. That can create thousands of pages without producing a clear authorship record.

A useful file should make the human role visible. It should help answer:

  • Who originated the premise, argument, characters or framework?
  • Who decided what the book was trying to accomplish?
  • Who designed and revised the structure?
  • Which generated suggestions were rejected?
  • Which passages were substantially rewritten?
  • Who verified the claims and sources?
  • Who controlled the final expression and approved publication?

Prompts can support those answers, but prompts alone are not authorship. The stronger evidence is the pattern of human direction, selection, revision and responsibility across the project.

This Story Also Raises Questions About Fairness

A manuscript record is a sensible professional habit. It must not become an excuse for unequal scrutiny.

Falade has argued that Black authors receiving significant attention have faced suspicion that their success could not be their own. Whether every comparison is equivalent or not, the concern deserves serious treatment. Documentation standards can become unfair if unknown, marginalized or less technically organized writers are expected to prove themselves in ways established authors are not.

There are also practical limits:

  • Writers do not all work in the same way.
  • Some draft by hand, overwrite files or develop stories orally.
  • Version histories can be incomplete.
  • AI detectors can produce false positives and inconsistent results; a score should not be treated as proof that a specific manuscript was AI-written.
  • Proof that AI was not used may be impossible to make absolute.
  • Contracts often fail to define what kinds of AI assistance are permitted.

The publishing industry needs clearer policies, consistent expectations and procedures that do not treat suspicion as proof. Authors, meanwhile, should protect themselves by making their own process easier to explain.

The Jack Righteous Position: Use AI Without Surrendering Direction

The lesson is not that responsible authors must avoid every AI tool. Spellcheckers, transcription tools, research systems, editing software and generative assistants now overlap in complicated ways.

The better standard is human-led creation:

  • Use AI for a defined purpose.
  • Know what the tool contributed.
  • Keep the important stages of the work.
  • Rewrite and decide rather than accepting output passively.
  • Verify sources, quotations and factual claims.
  • Disclose AI-generated material where contracts or platforms require it.
  • Remain accountable for the final book.

AI should not make the author’s role harder to see. Used responsibly, it should make human direction more intentional.

The 7-File Starter Record Every Author Can Build Today

You do not need a complicated legal archive to begin. Create one project folder and add these seven files or subfolders.

  1. 01 — Original Concept: Save the earliest idea, purpose, premise, reader promise, voice note or handwritten scan.
  2. 02 — Human Outline: Preserve the structure you created or approved before full drafting.
  3. 03 — AI-Use Notes: List the tools used, what each tool did and whether generated material entered the book.
  4. 04 — Major Drafts: Save milestone versions rather than endlessly overwriting one file.
  5. 05 — Human Revision Record: Keep selected before-and-after passages, major decisions and editorial changes.
  6. 06 — Sources and Permissions: Track research, quotes, images, licences, contributors and uncertain rights questions.
  7. 07 — Final Manuscript and Approval: Save the final version, approval date, publication destination and disclosure notes.

This starter record does not prove a negative, guarantee copyright protection or compel a publisher to accept a manuscript. It is still more useful than retaining only the final file.

A starting folder structure you can copy

BOOK_PROJECT/
  01_Original_Concept/
  02_Human_Outline/
  03_AI_Use_Notes.txt
  04_Milestone_Drafts/
  05_Human_Revisions/
  06_Sources_Permissions/
  07_Final_Approval/

For an older book, use the earliest authentic notes you still have and create a dated retrospective summary for anything reconstructed from memory. Mark gaps rather than inventing evidence. Keep private drafts, collaborator messages and sensitive research secure; you do not have to upload them to public AI-detection websites to create a useful record.

What Authors Should Ask Before Signing or Submitting

As AI clauses become more common, authors should not wait until a dispute erupts to learn what an agent, publisher, client or platform expects.

  • How does the agreement define generative AI?
  • Is brainstorming treated differently from generated final text?
  • Are editing, translation, transcription or image tools covered?
  • What must be disclosed, and at what stage?
  • What representations is the author making about authorship and originality?
  • What records could be requested if questions arise?
  • What happens if the parties disagree?

For high-value deals, professional contract advice matters. A general internet checklist cannot replace review of the actual agreement.

The Bigger Shift in Publishing

The Falade dispute arrived during a year in which publishing has been confronting AI from several directions at once. Authors have faced accusations about generated text. Publishers have debated disclosure and detection. Major rights holders have pursued claims concerning the use of books to train AI systems. Observational research has also studied the frequency of detected AI-generated text in self-published genre books. Those results depend on detection methods and sampled titles and are not evidence about this author or his manuscript.

The pressure is not going away. Authors will increasingly need two forms of readiness:

Creative readiness

A manuscript with a clear voice, strong structure, useful ideas and meaningful human judgment.

Process readiness

A record that explains the tools, drafts, sources, revisions, permissions and decisions behind that manuscript.

One cannot replace the other. A well-documented weak book is still weak. A brilliant book with a disputed and unexplainable creation history may now carry risks that writers did not previously anticipate.

What Every Author Should Learn From This Case

The later September interview provides more of Falade’s account without settling the authorship dispute. His denial of AI-written prose and his concerns about racial and cultural bias remain essential to a fair account. The reviewed reporting does not establish that AI wrote his novel.

But the collapse of a reported multimillion-dollar opportunity demonstrates something every author can act on now:

Do not wait until someone questions your work to reconstruct how it was created.

Save the origin. Preserve milestone drafts. Record important AI use. Track sources and permissions. Keep editorial history. Make the human decisions visible.

Your manuscript deserves more than a final file. It deserves a creation record you can understand, maintain and stand behind.

Continue the Author Path

Learn what to document before you publish

This article introduces the risk. The next guide explains the documentation system: human authorship, AI-generated versus AI-assisted material, prompt and draft history, sources, images, rights notes and publication records.

Read AI Copyright for AuthorsOpen the 12-Step APPLY Workflow

Continue free · Optional tools for different writing needs

Build your own manuscript record without a purchase.

The seven-file method and connected LEARN and APPLY lessons are free. These resources help with author planning and writing—not with proving that a publisher must accept a manuscript.

FREE DOWNLOAD · C$0

Audience, Point of View & Message Student Pack

Editable PDFs for establishing your original reader promise and author perspective.

Get the free author pack →
OPTIONAL · C$5 ONE-TIME

Righteous AI Content Engine

Workbook for human-led writing, revisions and publishing checks—not a copyright or provenance certificate.

Explore the writing workbook →

Sources and Further Reading

The Guardian: “$2m crime novel deal collapses amid questions over AI use” — primary reporting used for the deal, withdrawal, agency statements, Falade’s denial and his allegation of racial bias.

The Independent, September 16, 2026 — Falade’s later account of AI use for research, Yorùbá narrative traditions and retained manuscript records.

Publishers Weekly, August 4, 2026 — further industry reporting on the withdrawal and the author’s response.

Chakrabarty, Liu, Ginsburg and Dhillon: “Generative AI floods and dilutes the market for books” — Observational July 2026 research using AI detection on self-published genre fiction; it cannot establish facts about Falade.

Jack Righteous: AI Copyright for Authors — practical documentation guidance for AI-assisted writing.

Educational notice: This article provides publishing-process education, not legal advice. Copyright rules, publisher contracts, platform policies and AI-disclosure requirements can change. Review the current terms that apply to your project and seek qualified advice where commercial or legal risk is significant.

Zurück zum Blog

Hinterlasse einen Kommentar

Bitte beachte, dass Kommentare vor der Veröffentlichung freigegeben werden müssen.

articleall levelsHow to Use Jack Righteous
On this page

    Keep Jack Righteous in your Google results

    Make Jack Righteous a preferred source.

    Google can highlight preferred publications more prominently for you in Top Stories, AI Mode and AI Overviews when those features are available.

    The Righteous Beat

    Get useful writing, publishing and creator-platform guidance.

    Join the free newsletter →