> ## Content Index > Fetch the complete content index at: https://modalpathethics.com/llms.txt > Use this file to discover other available public pages before exploring further. # Applied Case: Call Me, I'll Hide The Body - URL: https://modalpathethics.com/applied-case-call-me-ill-hide-the-body/ - Published: 2026-08-03T10:30:05.000Z - Updated: 2026-08-03T10:30:04.000Z - Description: The manuscript passed. Its path became the case. - Author: Aidan Lawson - Tags: Applied Case, Modal Path Ethics, Modal Systems This manuscript did not change. It remained the same crime novel that had reportedly triggered a fourteen-way American auction, several six-figure British bids, and an offer above **two million dollars**. Then, one question entered the room. > **How did this manuscript come into existence?** The sentences remained where they were. The plot did not collapse. The characters did not become less alive. Editors did not collectively discover that they had secretly hated the book all along. Its reachable future collapsed anyway. According to [the Guardian's account](https://www.theguardian.com/books/2026/jul/31/crime-novel-deal-collapses-questions-ai-jerry-falade-call-me-ill-hide-the-body?ref=modalpathethics.com), Jerry Falade's debut novel, *Call Me, I'll Hide the Body*, had become one of those publishing events designed to make every participant feel that history had briefly entered the conference call. The book was headed toward a major deal and a planned 2028 publication. Then Falade's agents withdrew it from sale after saying they could no longer authenticate how the manuscript had developed from “origin to completion.” Falade denies using generative artificial intelligence. He says the collapse was driven by racial bias. His agents say concerns about the manuscript's provenance intensified after aspects of his account changed. One of them still described the novel as extraordinary, “however it came to be drafted.” The public record currently establishes a dispute. It does not establish a verdict. Modal Path Ethics will not manufacture one even though a clean answer would make the article move faster. Every possible answer leads somewhere different. - If Falade wrote the manuscript without generative artificial intelligence, - a human-authored debut may have had a multimillion-dollar future stripped away by **suspicion**. - If Falade substantially used generated prose and denied it, - his agents may have faced a serious problem of candor, rights, and contractual warranty. - If he used artificial intelligence somewhere inside the large and poorly mapped middle between spellcheck and synthetic drafting, - the industry may have attempted to govern a gradient with a purity test. Those are three different fields. They must remain different. They still converge on one institutional fact: > **Publishing authenticated the product before it could authenticate the path.** That is the case here. --- ## **The Book Was Never Only the Book.** Publishing occasionally imagines itself as the serious wing of entertainment. It has editors, paragraphs, imprints, literary prizes, and people who can identify a semicolon in the wild. It also has auctions, star-making, publicity cycles, adaptation rights, prestige markets, speculative acquisition, and the old entertainment-industry talent for converting one human being into a saleable story about where the work came from. A publisher buying a debut novel is not buying a sequence of sentences alone. It is buying a bundle. - A manuscript. - A chain of rights. - A warranty of originality. - A claim of human expressive control. - An author who can revise, answer questions, participate in publicity, and write the next book. - A public identity through which the work can be introduced to readers. - A relationship of trust among writer, agent, editor, publisher, bookseller, reviewer, and audience. The book belongs at the center of that bundle. The rest remains load-bearing. This is easy to dismiss as publishing hypocrisy. Some of it may become hypocrisy very quickly. The deeper structure is legitimate. A reader may care whether a novel was written by another human being. That preference can concern labor, encounter, artistic intention, historical position, voice, or the sense that a book is one human consciousness making itself available to another through form. The reader does not need to prove that machine-assisted prose is aesthetically impossible before valuing human authorship as a characteristic of the work. A publisher also has legal and contractual reasons to care. The [United States Copyright Office](https://www.copyright.gov/ai/?ref=modalpathethics.com) continues to distinguish human-authored expression from material whose expressive elements were determined by a machine. Human selection, arrangement, and revision may receive protection where they contain sufficient human authorship. Machine-generated material itself does not become copyrightable because a person wanted it very badly. The [Authors Guild's current guidance](https://authorsguild.org/resource/ai-best-practices-for-authors/?ref=modalpathethics.com) likewise warns that generated text can collide with originality warranties, copyright registration, disclosure duties, and reader trust. Its guidance distinguishes research, ideation, editing, and the direct incorporation of generated prose because those transitions do different work in the manuscript. Human authorship is therefore a legitimate product characteristic. The institutional failure begins when an industry promises that characteristic without possessing an adequate instrument for verifying it. The Falade case became explosive because the prose retained full value after the account of its origin lost institutional trust. This manuscript remained desirable as an object. It became unreachable as a commercial relation. That tells us exactly where the industry located the breach. It was not in the quality of the endpoint. It was in the path claim attached to the endpoint. --- ## **Authorship Is a Path Claim.** Authorship is often treated as a label attached to a finished object. > **Written by Jerry Falade.** Place the name under the title. Put the photograph on the jacket. Add a biography. Send the author to a festival where someone will ask how the idea came to him and whether he always knew the ending. The label looks static. The claim beneath it is historical. **Authorship is a path claim.** It says something about how the present object arrived. It does not say the writer used no instruments. Writers have always used instruments: dictionaries, notebooks, research assistants, editors, style guides, search engines, transcription software, spellcheck, workshops, friends, archives, index cards, stolen afternoons, legal pads, old letters, and the deeply advanced literary technology known as asking someone whether a sentence sounds stupid. A serious authorship claim instead concerns control and contribution. - Who determined the expressive content? - Who made the structural decisions? - Which passages entered from another source? - Which tools proposed language rather than correcting it? - What did the writer accept, reject, rewrite, or generate independently? - What role did editors or collaborators play? - What rights attach to the result? - What was represented to the agent and publisher? No final manuscript contains a perfect answer to those questions. The endpoint carries traces of its production history, but it does not contain a full audit log. A clean paragraph does not announce whether it emerged through six months of revision, one strange morning, dictated speech, developmental editing, generated alternatives, or some sequence involving all of them. A strong scene does not arrive with a little provenance certificate stapled to its ankle. This has always been true. Ghostwriting, plagiarism, undeclared collaboration, editorial overreach, translation, book packaging, and fabricated memoir have existed for a long time. Publishing has **never** possessed transparent access to the entire production path. Generative artificial intelligence changes the scale and the ambiguity. It can produce completion-shaped language at speed. It can enter anywhere from research through drafting. It can be used once, continuously, visibly, invisibly, responsibly, lazily, deceptively, or as an accessibility instrument. It can supply a phrase, a scene, an outline, a diagnostic response, a rewrite, a false fact, a useful objection, a whole chapter, or an enormous quantity of prose that a human later reconstructs. Those relations do not collapse into one event called **AI use**. The Authors Guild learned this while building its own [Human Authored certification](https://authorsguild.org/news/clarifications-regarding-the-human-authored-certification-terms/?ref=modalpathethics.com). In May 2026, it revised the definition after writers objected that categories such as brainstorming and outlining reached too broadly. The current standard focuses on the text itself and permits a de minimis amount of machine-generated or machine-modified material for functions such as spelling, grammar, or indexing. The organization constructing one of the clearest human-authorship standards in publishing had to correct its categories after first contact with writers' actual practices. That is healthy. It is also evidence that this boundary is being built in real time. Yet much of the industry still reaches for one easy question: > **Did you use AI?** That sentence is trying to ask too much. That one is trying to ask about process, degree, expressive control, retained generated text, legal rights, disclosure, labor, ethics, contract, reader expectation, and candor in five words. The result is a question with the emotional force of a moral test and the technical precision of a smoke alarm in a kitchen. --- ## **Three Possible Fields.** The Falade case cannot be audited honestly without preserving all three live fields. ### **Falade did not use generative artificial intelligence.** In this field, a human-authored crime novel passed readers, editors, agents, acquisitions departments, and a competitive market. Then, suspicion about an invisible production history became strong enough to close the path. The injury here would be severe. A debut author cannot easily prove **the non-occurrence of a hidden act**. Drafts can help. Notes can help. Version histories can help. Correspondence can help. Prior writing can help. None offers perfect negative proof. A writer may overwrite files. A writer may compose in one document. A writer may dictate. A writer may work longhand. A writer may delete embarrassing drafts, use accessibility software, draft in fragments, rebuild from memory, or produce cleaner early prose than suspicious observers believe a debut author should be able to produce. The burden now becomes structurally ugly: > **Prove that no forbidden transition occurred anywhere in the path.** A missing record can then become evidence. A nervous answer can become evidence. A revised answer can become evidence. The demand for proof like this can produce the very inconsistencies **later treated as proof of concealment**. If **this** is the real field, the scandal is an epistemic execution conducted under commercial urgency. The manuscript was **accepted as** **art** and rejected as *biography*. ### **Falade substantially used generated text and falsely denied it.** In this field, the agency had fully legitimate reasons to withdraw. A publisher facing substantial undisclosed generated prose may confront copyright uncertainty, an inaccurate originality warranty, editorial difficulty, reputational exposure, reader deception, and a broken relationship with the author. An agent does not have to carry a multimillion-dollar representation into contract while believing the central process claim may be false. Candor matters. The [Trespass Machine](https://modalpathethics.com/applied-case-the-trespass-machine/) also remains active. Writers and artists are responding to real extraction, substitution, impersonation, deskilling, uncompensated training, and forced participation. A human-authorship label has material significance inside an industry where human creative labor is already being used to build systems capable of replacing parts of that labor. If substantial undisclosed generated prose entered the manuscript, the answer cannot be that the novel was good, so the path no longer matters. It matters. Yet this field still produces an institutional indictment. Fourteen bidders could recognize an exciting commercial object. They could price it, fight over it, imagine its audience, and prepare its future. Their ordinary acquisition practices could **not** establish whether the object possessed the provenance they believed they were purchasing. - Professional taste passed the manuscript. - Professional taste could not possibly authenticate the manuscript. Those are different instruments. ### **Falade used artificial intelligence somewhere inside the middle region.** This field is likely to become the ordinary future of authorship disputes. - A writer researches with a model. - A writer asks for twenty possible motives and uses none of the phrasing. - A writer asks for structural criticism. - A writer receives a generated outline and changes it beyond recognition. - A writer pastes a paragraph into a tool for line-level alternatives. - A writer retains three sentences. - A writer uses synthetic transcription. - A writer uses a grammar system whose current product quietly performs more generative rewriting than its old product did. - A writer with a disability uses assistance that another writer would call automation. Where, **exactly**, does authorship change category? The answer depends on the transition. Research is not retained prose. Ideation is not sentence determination. Sentence determination is not whole-work determination. A generated paragraph extensively rewritten by a human presents a different rights and authorship problem from a generated chapter lightly edited for continuity. Undisclosed use presents a different trust problem from disclosed use accepted under contract. The industry cannot govern this field through a binary contamination category. It needs a process taxonomy. Every immediate ruling changes across these three fields. All three still converge here: > **Publishing does not yet possess a mature authorship-provenance instrument.** --- ## **The Completion Engine Passed.** [The Completion Engine](https://modalpathethics.com/applied-the-completion-engine/) described one of the strongest anti-artificial-intelligence pressures before this case arrived. Generative systems can make unfinished things look done. They can smooth roughness, bridge missing steps, imitate confidence, and convert a partial structure into an artifact with the outward posture of completion. The danger is larger than bad prose. A culture exposed to constant machine completion can begin treating friction as incompetence and incomplete thought as a defect that should have been removed before anyone had to encounter it. The Falade case reveals another edge. > **Completion can conceal path.** This statement does not establish that Falade used a model. It explains why the finished manuscript cannot settle the accusation. A polished novel has **always** hidden most of its workshop. Generative artificial intelligence adds a new production route whose output can resemble ordinary completion closely enough that the surface cannot reliably identify the path beneath it. Professional readers are trained to answer questions about the **endpoint**. - Does the scene work? - Does the voice hold? - Does the structure carry tension? - Can the book reach an audience? - Can the ending pay its debts? They are not trained forensic archaeologists of every sentence's origin. Nor should literary judgment become a fake detector. The cultural search for machine “tells” has already produced an incredibly unstable folk science: em dashes, symmetry, certain transitions, tidy abstractions, repeated sentence rhythms, familiar metaphors, excessive polish, insufficient mess. All of those features existed in human writing before the first chatbot acquired a login screen. Human writers also read generated prose, react against it, imitate it accidentally, and share stylistic pressures with the data from which models learned. Research on detection has repeatedly found very serious limits. Studies have shown [bias against non-native English writing](https://doi.org/10.1016/j.patter.2023.100779?ref=modalpathethics.com), vulnerability to [paraphrasing and spoofing](https://arxiv.org/abs/2303.11156?ref=modalpathethics.com), and [broad reliability problems across commercial tools](https://doi.org/10.1007/s40979-023-00146-z?ref=modalpathethics.com). A detector can contribute one weak signal inside a larger inquiry. It cannot serve as a synthetic judge wearing a lab coat. The most dangerous move is to let *taste* impersonate forensics. > **This prose feels too finished for this person.** That sentence can contain real editorial intuition. It can also contain class, race, linguistic expectation, prestige bias, unfamiliarity with the writer, or the simple shock of encountering talent before an institution has produced a credential for it. The Completion Engine makes provenance harder. It does not make suspicious confidence smarter. --- ## **Total Faith != an Instrument.** Falade's agent described representation as requiring “total faith.” That makes sense at one level. An agent signs a writer's warranties, reputation, future labor, and account of the work into a chain of relationships. If the agent comes to believe the author is changing a material story, the relationship may become impossible to continue. Trust is real infrastructure. It is still not a provenance procedure. The public account suggests a system with two stable settings: - total faith; - total withdrawal. Between those settings lies the entire missing instrument. There was apparently no standard independent authorship review waiting to receive the dispute. No public evidentiary category distinguished unresolved concern from established deception. No shared protocol told publishers which process records were relevant, which questions were legitimate, how much inconsistency mattered, or what remedy belonged to each possible finding. This does not prove the agency acted wrongly. The agents knew details the public does not. They may have reached the prudent decision under the facts available to them. The larger architecture remains poor. A fourteen-way auction is a dramatic way to discover that your provenance procedure was a conversation. Auction pressure likely sharpens the problem. Competitive acquisition rewards speed, confidence, enthusiasm, and fear of losing the next major book. The manuscript can become a market event before the industry has completed the slower work of establishing what kind of object it is buying. Quality is authenticated at pace because quality drives the bid. Production history remains inside personal trust because production history is assumed to be settled until somebody breaks the assumption. Then suspicion enters. The same speed runs backward. Publishers pause. Agents withdraw. Announcements disappear. The person who was a discovery becomes an exposure. The book that everyone needed to acquire becomes the book nobody can safely touch. The entertainment industry knows how to accelerate desire and risk. It remains **much** worse at building a middle procedure between them. --- ## **Suspicion Enters the Manuscript.** [Field Instruments: Active Information](https://modalpathethics.com/field-instruments-active-information/) supplies the deeper structure. Information becomes active when its uptake changes reachability. The artificial-intelligence concern did not sit beside the manuscript as one additional fact waiting for later review. It entered the commercial field and reorganized the manuscript's meaning. The same polished passage could now be read as evidence of machine generation. The same rapid drafting history could become evidence. The same lack of conventional drafting artifacts could become evidence. The same authorial confidence could become overconfidence. The same confusion could become evasion. The same changed explanation could become proof that every prior explanation was false. The classification changed the classified field. Falade was no longer only an author presenting a novel. He had become the person required to authenticate the absence of a disputed production path under rapidly increasing commercial pressure. According to Falade's account, rumours spread and the deal structure collapsed within hours. That claim has not been independently established in every detail. The speed itself belongs to the audit because speed changes what kinds of correction remain available. A concern handled privately can still be investigated. A concern circulating through publishers, trade reporting, social media, and public statements becomes epistemic weather. Every later fact arrives through it. Active Information asks a brutal question: > **Which person acquired the burden of disproving the new default?** In this case, the answer appears clear. The publishing industry did not need to prove machine generation before the suspicion became commercially decisive. Falade needed to restore enough confidence for every participant to keep accepting risk. That asymmetry may be unavoidable in some contractual relationships. A publisher is not a criminal court. An agent is not required to reach certainty beyond reasonable doubt before ending representation. The absence of criminal procedure does not authorize procedural emptiness. A high-stakes creative market still needs distinctions among: - a rumour; - a stylistic concern; - an unexplained inconsistency; - weak technical evidence; - strong textual evidence; - documented generated text; - an inaccurate disclosure; - deliberate contractual misrepresentation. Without those distinctions, *suspicion* does all the routing. The field decides first. The evidence arrives later to explain why the field had to decide. --- ## **The Racialized Credibility Field.** Falade alleges racial bias. He argues that Black writers receiving major deals have been subjected to disproportionate suspicion that their work could not be their own. He points toward other recent authors whose books were challenged or withdrawn after artificial-intelligence allegations. The available public record does not establish the full pattern. One disputed case cannot establish it. One disputed case also cannot close the inquiry. The question is empirical: - Who is asked to prove a drafting history? - Whose unusual polish is interpreted as talent? - Whose unusual polish is interpreted as machine assistance? - Who receives a confidential inquiry? - Who becomes a public controversy? - Whose explanation is treated as normal process variation? - Whose explanation is treated as instability? - Who receives time, counsel, and an independent review before a deal collapses? - Who remains represented while the facts are investigated? - Which writers are expected to possess a neat archive before anyone believes they could have written the book? Publishing has no reason to assume its suspicions arrive outside history. Credibility has always been unevenly distributed. Some writers enter the room already carrying institutional confidence. Their education, agent, career, class position, accent, prior publications, references, and social familiarity make authorship feel unsurprising. Others arrive as discoveries, which is the industry's flattering word for people it had not prepared itself to recognize. A provenance system built from intuition, rumour, stylistic expectation, personal credibility, and total faith will inherit the field's existing credibility priors. Technical instruments do not automatically solve that. The study finding that detectors disproportionately flagged non-native English writing matters because it shows how a system can transform linguistic difference into suspicion while presenting the result as neutral measurement. None of this proves that racial bias caused the Falade decision. It proves that the industry cannot answer the allegation by insisting that everyone involved had good intentions. “Good intention” does not audit distribution. The repair is data. Agencies and publishers should record disputed-authorship inquiries in an anonymized audit system: what triggered them, what evidence was requested, how long the process lasted, what standard was used, what outcome followed, and whether patterns appear across race, nationality, language background, disability, class, genre, prestige, or route to publication. If the pattern does not exist, the audit can show that. If it does exist, the industry will finally have something stronger than reassurance with which to confront it. --- ## **Provenance and Purity.** The strongest defense of human creators cannot be surrendered here. Training on uncompensated work matters. Substitution matters. Synthetic imitation matters. Translation, narration, illustration, and editorial jobs matter. Consent matters. Disclosure matters. A publisher's desire to reduce cost does not become artistic progress because a machine can produce an acceptable surface. [The Trespass Machine](https://modalpathethics.com/applied-case-the-trespass-machine/) named those injuries directly. [Applied Case: The Anti-AI Religion](https://modalpathethics.com/bad-religion-the-anti-ai-religion/) named the next failure. A damaged field can produce a purity system. The phrase **AI touched it** then begins doing the work that process analysis should have done. It collapses research, grammar correction, transcription, brainstorming, developmental feedback, sentence generation, sustained drafting, plagiarism, undisclosed use, accessibility, displacement, rights, and deception into one sign of contamination. That sign is emotionally efficient. It is institutionally useless. Provenance asks: - What happened? - Which transitions mattered? - What entered the final manuscript? - Who retained expressive control? - What was disclosed? - What rights and obligations follow? Purity asks who is clean. Publishing needs provenance. It keeps reaching for purity because purity is so much easier to market. The asymmetry becomes especially sharp when the publishing house is already hybrid. Publishers are experimenting with generated cover imagery, marketing copy, translation, editing tools, synthetic narration, metadata, and internal manuscript processing. The [Authors Guild's model contract clauses](https://authorsguild.org/advocacy/artificial-intelligence/ai-model-clauses/?ref=modalpathethics.com) now address this from both directions: authors disclose generated material, while publishers should seek consent before using artificial intelligence for narration, translation, artwork, substantive editing, or manuscript ingestion. That symmetry is correct. - **The author should disclose which instruments entered the manuscript.** - **The publisher should disclose which instruments enter the book.** A company cannot sell “human creative sanctity” at the front of the building while automating narrators, translators, illustrators, editors, and promotional artists through the loading dock. --- ## **No Creative Panopticon.** The obvious response to uncertain provenance can become its own industry disaster. Publishers may begin demanding complete version histories, prompt logs, screen recordings, cloud records, keystroke histories, notebooks, metadata, device access, and a permanent archive of every creative transition. This would make authorship easier to police by turning writing into supervised labor. This is a terrible idea. A writer should **not** need to produce a keystroke alibi. Creative process contains privacy for good reason. Drafts can include autobiography, research into violence, abandoned political thought, private correspondence, ugly experiments, medical material, sexual material, legal risk, other people's stories, and sentences whose only moral accomplishment was being deleted before sunrise. Writers also work differently. Some preserve every draft. Some overwrite. Some dictate. Some compose by hand. Some work through fragmented notes. Some build a clean draft internally before writing. Some rely on editors. Some use accessibility instruments. Some destroy material because they do not want every failed attempt retained for future discovery, litigation, gossip, or machine training. A universal surveillance requirement would distribute burden unevenly and reward whichever creative life already resembles enterprise document management. The right repair is **bounded provenance**. A provenance instrument should collect enough information to govern authorship, rights, and candor without claiming ownership over the writer's entire path. That can include: - a standardized process disclosure; - clear definitions supplied before submission; - voluntary retention of representative drafting materials; - source and collaboration notes where relevant; - a confidential review process for disputes; - contractual warranties matched to the actual categories; - narrow requests for additional evidence when a specific concern arises. The request should be proportional to the concern. The inquiry should preserve privacy beyond the disputed question. The author should know what standard is being applied. The institution should not convert the absence of corporate-style telemetry into evidence of guilt. Provenance must be strong enough to protect readers and creative workers. It must remain weak enough to leave a writer with a private mind. --- ## **Build the Authorship Instrument.** The repair is available. It will require publishing to replace one dramatic question with several boring ones. This is excellent news. Boring questions are among civilization's better defenses against the sudden arrival of moral weather. ### **1\. Define the categories before submission.** Every agency and publisher should publish a process taxonomy. At minimum, it should distinguish: - ordinary software assistance; - artificial-intelligence-assisted research; - ideation and structural feedback; - transcription and accessibility support; - sentence-level modification; - generated text substantially rewritten by the author; - generated text retained in the final manuscript; - sustained machine-generated drafting; - prohibited undisclosed use. The definitions should state which uses are permitted, which require disclosure, which require contract approval, and which are incompatible with a human-authorship label. A writer should not discover the ontology during the investigation. ### **2\. Separate the four inquiries.** The Falade case became combustible because four questions could collapse into one. 1. **Quality:** Is the manuscript good? 2. **Authorship:** Who determined its expressive content? 3. **Rights:** What can be owned, licensed, warranted, and defended? 4. **Candor:** Did the author accurately describe the process? A manuscript can pass one inquiry and fail another. A beautiful book can carry a defective rights chain. A substantially human-authored book can include a disclosure failure. A contractually compliant book can still be bad. A writer can use a permitted tool and describe it badly under pressure. Each question requires different evidence and different remedies. ### **3\. Use a one-page process disclosure.** The default instrument should be small. A standard disclosure can ask which tool categories entered the work, whether generated language remains in the submitted text, whether any third party contributed undisclosed prose, and whether the author can provide representative process evidence if a material dispute arises. This is not a confession. It is production metadata. The disclosure should travel with the manuscript through agency, auction, contract, editing, and rights sale so the same process claim does not have to be reinvented in every room. ### **4\. Build a disputed-authorship procedure.** When a serious concern arises, the author should receive the specific concern rather than a cloud of suspicion. The process should include: - written notice of the issue; - a meaningful opportunity to answer; - a defined evidentiary standard; - confidential independent review where the stakes are high; - qualified textual, technical, contractual, and editorial analysis; - a rule that detection software cannot serve as sole adjudicator; - an appeal or second-review path; - public language matched to what the evidence actually establishes. The finding should distinguish **unresolved provenance** from **substantiated generated text** and both from **demonstrated misrepresentation**. Those labels are not interchangeable. ### **5\. Make remedies proportional.** A grammar tool, a generated sentence, a machine-developed outline, an undisclosed synthetic draft, and deliberate contractual deception are different events. The available verbs should include: - clarify; - disclose; - document; - revise; - remove; - relicense; - amend the warranty; - renegotiate; - withdraw; - terminate. Publishing currently behaves as though the main verbs are **believe** and **destroy**. That is poor instrument design. A middle procedure protects authors from catastrophic false suspicion and protects publishers from being forced to ignore material breaches because the only alternative is total collapse. ### **6\. Audit suspicion itself.** Every provenance system must examine how its own alarms are distributed. Agencies and publishers should conduct periodic independent audits of who gets investigated, what triggers investigation, what evidence is treated as sufficient, and which outcomes follow. The audit should include linguistic and demographic disparities, because an authorship instrument that repeatedly converts unfamiliar voice into machine suspicion has become a distortion field. The audit should also examine prestige. An established writer with a long archive may receive interpretive generosity. A debut author may receive a demand for proof. A famous author may be allowed to call generated drafting experimentation. An unknown author may be treated as counterfeit. “Standards” that change with market power are publicity rules wearing ill-fitting ethical clothes. ### **7\. Make publisher provenance symmetrical.** The publisher should provide a matching disclosure for the finished book. Did artificial intelligence enter substantive editing, translation, narration, cover art, marketing, metadata, publicity, or adaptation materials? Was the author's manuscript uploaded into a third-party model? Were the author and other creative workers asked for consent? What rights were retained? What human jobs were displaced? The book's path continues after acquisition. Authorship provenance does not end at the writer's desk. --- ## **The Entertainment Industry Recognizes Value Faster Than Responsibility.** [Applied Case: The Missing Link](https://modalpathethics.com/applied-case-the-missing-link/) audited an entertainment industry that changed its classification of Joe Martin according to managerial need. The moral circumstances are profoundly different. The cases should never be equated. The structural rhyme remains useful. A creative industry can recognize value with extraordinary speed while delaying recognition of the relation that produced the value. In *The Missing Link*, the same living being could become actor, celebrity, child, beast, or danger as each classification served the room. Here, the manuscript moved through another sequence. - Genius. - Discovery. - Auction object. - Rights package. - Provenance concern. - Commercial exposure. The artifact did not move between categories by changing its sentences. The institution changed the category around it. This is what entertainment machines do especially well. They build fast channels for desire. They can identify what audiences may love, what editors want, what rivals fear losing, what publicity can amplify, and what money should move before anyone else reaches the property. Responsibility arrives through slower channels. Who made this? What exactly are we buying? Which labor is being protected? Which rights exist? Which person now carries the cost of our uncertainty? What procedure survives after confidence fails? The industry becomes dangerous where its market instruments are more mature than its contact instruments. The Falade manuscript reportedly passed every test publishing knew how to administer to the book. Then publishing discovered that the test it needed concerned the path. That test had not been built. --- ## **The Ruling.** A publisher is allowed to care who wrote the book. A reader is allowed to care whether the work's expressive language was substantially authored by another human being. An agent is allowed to withdraw when a material warranty can no longer be trusted. A writer is allowed to reject a public verdict that the available evidence has not established. None of those permissions converts suspicion into proof. None makes artificial-intelligence use one clean category. None gives an entertainment industry the right to demand a writer's entire private process because **it failed to define its own standards before the auction**. The manuscript did not become bad. The path claim became commercially unstable. If Falade did not use generative artificial intelligence, the field may have punished a writer for being unable to prove a negative under conditions already reorganized by suspicion. If Falade substantially used generated prose and denied it, the field discovered a real breach after pricing the endpoint without authenticating what it was buying. If his process occupied the middle region, the field may have converted an undefined boundary into a binary judgment because binary judgments travel faster. Every version requires a different ruling. Every version indicts the instrument. > **Authorship is a path claim.** Build a system capable of distinguishing assistance, generation, expressive control, rights, disclosure, and candor. Build a dispute procedure that can investigate without declaring a public contamination event. Build an audit strong enough to detect whether suspicion follows race, language, class, disability, novelty, or prestige. Build symmetrical disclosure for the publisher's own machinery. Keep the provenance boundary bounded enough that the writer remains a person rather than a permanently monitored production surface. Until then, multimillion-dollar publishing will continue resting on an authorship system with two settings: - **Total faith.** - **Total withdrawal.** That is just not a mature standard.