Who Is the Author? The Dispute over Artificial Intelligence and Authorship

In September 2026, a novel regarded as one of the most successful books of the French literary season suddenly became the subject of a dispute over authentic authorship. C'était ça ou mourir by Thélyson Orélien, a Haitian-born writer living in Quebec, was on the first selection for the prestigious Goncourt Prize and had also won the Prix du Roman Fnac. But the publication of the results of a software analysis changed its fate.

On 21 September, an account on the social network X called "Balance ton Claude" claimed that the novel had been written almost entirely with artificial intelligence. The claim rested on an analysis by Pangram, a tool that assesses how likely a text is to have been produced by a machine.

Four days later, the Académie Goncourt removed the novel from its list of contenders. In its statement, besides the results of the AI-related analyses, the Académie cited cases of plagiarism in Orélien's earlier writings. The author accepted those cases and apologised, but on 13 October he wrote that no sentence of the novel had been made by a machine.

On 21 October the affair took a new turn. The Canadian publisher Boréal, after examining the working versions of the novel, announced that signs of AI use had been found in the metadata of some files. The publisher ended its collaboration with Orélien, but said it still did not know to what extent AI had been used and would probably never reach a definitive answer. The same day, Orélien told the newspaper Libération that he had written the novel himself, but accepted that he had used AI for research, including historical research. A week later, on 28 October, the jury of the Renaudot Prize also removed the novel from its final list.

This affair has confronted an old debate in literature with a new question. How can we know who the author of a work is?

A name that hid its author

In 1975 the Académie Goncourt awarded its prize to the novel La Vie devant soi (The Life Before Us) by Émile Ajar. Romain Gary, a writer who had also won the same prize in 1956 with the novel Les Racines du ciel (The Roots of Heaven), died on 2 December 1980. On 30 June 1981, Paul Pavlowitch, a young relative of Gary who for years had been the face of Ajar, revealed in a statement released by AFP that Ajar had been Gary's pen name. A few days later, on 17 July, a text Gary had written for publication after his death, Vie et mort d'Émile Ajar, was published and confirmed the secret.

By creating a literary identity and hiding his real name, Gary managed to win for a second time a prize that could be given to each writer only once. Critics had read and judged Ajar's books independently of Gary's fame and background.

The cases of Gary and Orélien are of two different kinds. Gary had hidden the author's identity, but there was no doubt that a human being had created the work. In Orélien's case, the question concerns the writing process itself. How much of the text was the result of the author's decisions, and how much may have come from a language-generation system?

Can software identify the author?

In July 2026, a study by Epoch AI, written by Jaeho Lee, examined the performance of three tools for detecting machine-generated text. The results showed that Pangram, in a set of 495 samples of human writing, did not wrongly identify a single one as machine-made.

Of the other two tools, GPTZero also did not wrongly identify any human text as machine-made, but Originality.ai wrongly identified nineteen samples as machine-made, and according to the report all nineteen had been written before ChatGPT appeared. In another test, when AI models imitated a writer's style after seeing samples of that writer's work, 30 of the 297 generated texts escaped detection by Pangram, 32 by GPTZero and 53 by Originality.ai. According to the report, the tools were weakest at detecting imitated scientific writing.

The study shows that the accuracy of detection tools depends on the type of text and on how it was produced. The report does not state the language of the texts, and all the samples were about 500 words long. The results of a single test set cannot be generalised to all literary works and all languages.

During the Orélien affair, the French writer Sophie Jomain said that Pangram had rated a book of hers, published in 2009, years before ChatGPT appeared, as 94 percent machine-made. This case points to possible errors by such tools, although on its own it cannot refute the other findings about Orélien's novel.

Software detection can therefore be one piece of evidence when examining the authenticity of a text, but judging authorship needs wider evidence, from early drafts and the process of rewriting to the author's own account of how the work took shape.

Who owns the copyright?

In the United States, the question of authorship of works made with AI has reached the courts.

On 2 March 2026, the US Supreme Court declined to hear the case of Stephen Thaler. Thaler had named his AI system as the sole author of an image and had sought to register its copyright. With his petition rejected, the appeals court ruling that authorship must be human remained in force.

The case of Jason Allen raises a more complex question. To create the image Théâtre D'opéra Spatial, Allen changed his prompts in Midjourney hundreds of times and then edited the image he chose. The US Copyright Office declined to register the work in the form he requested, and Allen sued in federal court in Colorado.

The main dispute is over how far human choices, repeated prompts and edits are enough to establish authorship. As of the latest legal report we found, from May 2026, the court had not issued a ruling in this case.

In its report of 29 January 2025, the US Copyright Office stated that using AI as a tool to assist human creativity, rather than to replace it, is no obstacle to copyright protection for a work. According to the same report, with current technology prompts alone do not give enough control for authorship. What is protected is the human contribution, that is, human expression that can be seen in the output, the creative selection and arrangement of material, and creative modifications to the output.

The rights of authors whose works were used

Legal disputes have not been limited to the ownership of AI outputs. On 20 July 2026, the federal court for the Northern District of California gave final approval to a 1.5 billion dollar settlement between the company Anthropic and a group of book rights holders.

According to the Authors Guild, the settlement concerned the unauthorised acquisition and copying of books from sources such as Library Genesis and Pirate Library Mirror. It covers claims about past copying, but it does not resolve every dispute about AI training or future outputs.

In this case, the question is not whether a machine can be an author. The question is the rights of people whose works were written by human beings and copied without permission.

A question beyond the author's name

The Goncourt affair, the court cases and the copyright disputes each reveal one side of the new relationship between literature and technology. The issue is not only finding a few machine-made sentences in a novel. The way a work is produced, the extent of human creative control, ownership of the text and the author's commitment to the reader all need to be reconsidered.

Writing novels has always involved using sources, being influenced by others, seeking advice and editing. The arrival of AI has made the line between helping a writer and producing part of a work more complicated.

Perhaps the question for literature today is no longer only whose name is printed on the cover of a book. We need to know how the text came into being and who has taken responsibility for its creative decisions.

Sources:
Livres Hebdo, 22 September 2026. https://www.livreshebdo.fr/article/thelyson-orelien-accuse-davoir-utilise-lia-je-ne-vois-pas-pourquoi-je-ferais-appel-une
ActuaLitté, 25 September 2026. https://actualitte.com/article/134194/prix-litteraires/le-prix-goncourt-exclut-le-roman-de-thelyson-orelien
Livres Hebdo, 25 September 2026. https://www.livreshebdo.fr/article/lacademie-goncourt-retire-thelyson-orelien-de-sa-selection
Livres Hebdo, 25 September 2026. https://www.livreshebdo.fr/article/thelyson-orelien-la-polemique-bouleverse-le-monde-du-livre
Paris Match/Belga, 13 October 2026. https://www.parismatch.be/culture/litterature/2026/10/13/thelyson-orelien-nie-tout-recours-a-lia-mais-reconnait-des-plagiats-je-le-regrette-profondement-HLMHXWDNMRFY5KGTLULCCF2GRI/
franceinfo, 21 October 2026. https://www.franceinfo.fr/culture/livres/affaire-thelyson-orelien/l-intelligence-artificielle-a-ete-utilisee-dans-l-ecriture-du-roman-de-thelyson-orelien-affirme-son-editeur-canadien-qui-met-fin-a-sa-relation-avec-l-ecrivain_8247700.html
Radio-Canada, 21 October 2026. https://ici.radio-canada.ca/nouvelle/2294673/boreal-editions-fin-collaboration-thelyson-orelien
franceinfo, 29 October 2026. https://www.franceinfo.fr/culture/livres/prix-litteraires/le-prix-renaudot-ecarte-l-ecrivain-thelyson-orelien-accuse-d-avoir-utilise-l-ia-de-la-liste-de-ses-finalistes_8258492.html
Académie Goncourt, list of winners. https://www.academiegoncourt.com/tous-les-laureats-prix-goncourt
Wikipedia, "Paul Pavlowitch". https://en.wikipedia.org/wiki/Paul_Pavlowitch
INA, Apostrophes, 3 July 1981. https://mediaclip.ina.fr/en/i05048016-paul-pavlowitch-and-emile-ajar.html
Epoch AI, Jaeho Lee, 15 July 2026. https://epoch.ai/data-insights/ai-detectors-false-negatives
Holland & Knight, 3 March 2026. https://www.hklaw.com/en/insights/publications/2026/03/the-final-word-supreme-court-refuses-to-hear-case-on-ai-authorship
Faegre Drinker, 28 May 2026. https://www.faegredrinker.com/en/insights/publications/2026/5/can-ai-prompting-satisfy-the-copyright-offices-human-authorship-requirement
U.S. Copyright Office, "Copyright and Artificial Intelligence, Part 2: Copyrightability", 29 January 2025. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
Authors Guild, 21 July 2026. https://authorsguild.org/news/court-grants-final-approval-anthropic-copyright-settlement/