हिंदी में पढ़ें —JantaScope हिंदी
AI NEWS

Did AI Write France’s Breakout Novel? Inside the Thélyson Orélien Controversy

Thélyson Orélien’s debut novel C’était ça ou mourir is at the centre of an AI-authorship controversy after a detector flagged passages as AI-generated. Orélien denies using AI, while plagiarism allegations and publisher concerns have intensified scrutiny, leading the Prix Goncourt to remove the novel from its longlist.

Did AI Write France’s Breakout Novel? Inside the Thélyson Orélien ControversyAI-generated editorial image | JantaScope

By Jeet Nirmal

Source: Le Monde, La Presse and Académie Goncourt; additional reporting and context from The Indian Express and Euronews. AI-authorship claims remain disputed by Thélyson Orélien.

From Literary Sensation to AI Controversy

For Thélyson Orélien, September 21 should have been one of the biggest days of his literary career.

The 38-year-old Haitian-Canadian writer had just won the 2026 Prix du Roman Fnac for his debut novel, C’était ça ou mourir (It Was Either That or Die), after the book emerged as one of France’s major literary successes of the season.

The novel had sold about 35,000 copies in five weeks in France and appeared on the lists for several prestigious French literary prizes.

But within hours of the Fnac victory, the conversation surrounding the book changed dramatically.

An initially little-known account on X called “Balance ton Claude” alleged that the novel had been written “almost entirely” using artificial intelligence. The name is a reference to Anthropic’s Claude AI and a play on the French #MeToo-era phrase “Balance ton porc.”

The accusation quickly spread across French social media and eventually became a much larger debate about authorship, AI detection and what evidence should be required before questioning a writer’s work.

What Exactly Was Alleged?

Balance ton Claude said it had submitted several passages from Orélien’s novel to Pangram, a tool designed to distinguish human writing from AI-generated text.

According to the account, the passages repeatedly returned results indicating “100% AI.” It therefore alleged that the novel was almost entirely generated using artificial intelligence.

The post became viral. The Indian Express reported on September 25 that it had accumulated more than 4.9 million views and 1,800 reposts by the time of publication.

Those results, however, are detector outputs and allegations—not proof of how the manuscript was actually produced.

That distinction has become central to the controversy.

Le Monde Conducted Its Own Test

The controversy gained additional weight when French newspaper Le Monde independently tested a substantial portion of the novel using Pangram.

According to Le Monde, the detector classified more than 95% of the text it tested as AI-written with a high confidence level.

As a comparison, the newspaper also submitted opening passages from around 15 other novels to the detector. Pangram classified those samples as human-written.

The result strengthened suspicions surrounding the book, but it still did not establish the novel’s authorship as a matter of fact.

AI-detection systems infer patterns from text. They do not reconstruct the writing process or provide a historical record showing whether an author typed a passage, generated it with an AI model, edited AI output or wrote independently.

Orélien Denies Using AI to Produce His Writing

Orélien has firmly rejected the central allegation.

In an interview with Canadian newspaper La Presse, he said:

“I have never used artificial intelligence to produce texts.”

He maintained that he could reproduce his writing style himself with a pen and attributed the unusual detector results potentially to the way he writes.

Orélien said his style deliberately relies on elements such as repetition and rhythm.

He has also pointed to the history of the project, saying the genesis of the book dates to 2017 and that a first version of the manuscript existed in 2019.

Those statements form part of Orélien’s defence; they do not independently resolve the technical question of how every part of the final manuscript was produced.

What Is C’était ça ou mourir About?

The novel follows Jonas, a young Haitian history and geography teacher who is forced to leave Haiti amid violence and insecurity and begins a journey across the Americas.

The book was first published in Canada before Grasset acquired the rights for French-speaking Europe.

Grasset published the French edition on August 19, 2026. Le Monde lists it at 272 pages and €21.50.

Before becoming a novelist, Orélien was primarily known for poetry.

The novel’s rise had been unusually rapid: strong sales, extensive critical attention and appearances on the initial selections for major awards including the Goncourt, Femina, Médicis and Renaudot. International translation rights were also being developed in more than 20 languages, according to reporting by Euronews.

Grasset Initially Defended the Author

Orélien’s French publisher, Grasset, strongly rejected the accusation when it first emerged.

Speaking to AFP, the publisher opposed what it described as malicious claims that the book had been fuelled or written by AI and said the accusations were deeply damaging to the writer.

The reaction reflected the uncertainty surrounding the evidence: an AI detector had produced striking results, but there was no independently established writing record proving that generative AI had produced the manuscript.

Canadian Publisher Pauses Promotional Activities

The controversy nevertheless began producing concrete consequences.

Canadian publisher Les Éditions du Boréal announced that it was temporarily suspending promotional activities surrounding Orélien and the book.

Boréal said it was working with the author to collect information that could help reconstruct the novel’s creation before the manuscript was submitted.

The publisher made an important distinction: it said no AI was used in its editorial work after receiving the manuscript, but said it was not yet able to confirm or disprove the allegations concerning the manuscript’s earlier creation.

Boréal cautioned against drawing premature conclusions.

Orélien’s planned participation in the Quebec City Book Fair was also cancelled amid the controversy.

Controversy Expands Beyond AI

The story became more complicated when questions emerged about some of Orélien’s older journalistic writing.

La Presse examined previous texts attributed to him, including a 2013 column that it said was almost entirely a lightly reworked translation of an article published days earlier by Philippine newspaper The Philippine Star.

These are separate plagiarism allegations concerning older writing and should not be treated as direct evidence that C’était ça ou mourir was written by AI.

Orélien has pushed back against the broader controversy, telling AFP there was a desire to “silence” him.

Prix Goncourt Removes the Novel From Its List

The most significant development came on September 25.

The Académie Goncourt removed C’était ça ou mourir from consideration for the Prix Goncourt, France’s most prominent literary prize.

The Academy said the decision was intended to protect the integrity of the prize and its role in recognising literature written by people. It cited the AI-authorship findings and plagiarism allegations in explaining its decision.

This is an institutional decision by the Goncourt Academy; it should not be confused with independent technical proof that AI wrote the novel.

The novel had been included in the initial Goncourt selection before the controversy erupted.

How Reliable Is Pangram?

This is where the story becomes more complicated.

Pangram is not simply a random online detector. Independent academic research cited by Le Monde found that it performed considerably better than several competing systems under the conditions tested.

A University of Maryland study estimated its false-positive rate at about 2%, while research from the University of Chicago placed it below 1%.

Those figures make its results relevant evidence worth examining.

They do not, however, make the detector infallible.

Researchers have warned that accuracy can change depending on the language, length and style of the text and the AI system potentially involved. Pangram itself has acknowledged errors in individual cases.

The challenge is particularly relevant here because Orélien’s novel is written in French and incorporates other languages when migrant characters speak.

Researcher Ophélie Fraisier-Vannier of the Toulouse Institute for Research in Computer Science told Le Monde that a human writer could potentially use linguistic patterns similar to those associated with a large language model. She also noted that unusual literary styles could complicate detection.

An AI Detector Cannot Reconstruct Who Wrote a Book

That limitation is perhaps the most important piece of context in the entire controversy.

AI detectors are classification systems. They analyse statistical characteristics associated with human and machine-generated writing and produce an assessment.

They cannot look backwards and show the actual process through which a paragraph was created.

Even a highly confident detector result cannot independently establish whether a passage was written entirely by AI, drafted by a person and edited using AI, generated and extensively rewritten, or simply written by a human whose style resembles patterns the detector associates with machine-generated prose.

Le Monde notes that identifying AI-generated text remains scientifically difficult and recalls that OpenAI abandoned its own AI-text classifier in 2023 because of accuracy limitations.

So, Did AI Write the Novel?

Based on the publicly available evidence as of September 25, 2026, there is no independently established answer.

What is confirmed is more specific.

Pangram has repeatedly flagged substantial portions of the novel as likely AI-generated, including in independent testing by Le Monde. The Goncourt Academy considered the concerns serious enough to remove the book from its competition. Meanwhile, plagiarism allegations concerning some of Orélien’s older writing have created additional scrutiny.

But Orélien categorically denies using AI to produce his texts, his Canadian publisher says it cannot currently confirm or refute the allegations concerning the manuscript’s creation, and the limitations of AI-detection technology prevent detector scores alone from providing definitive proof of authorship.

That unresolved gap—between unusually strong detector results and the limits of what those detectors can actually prove—is the central issue in the controversy.

For the publishing industry, the case may therefore have consequences extending beyond a single author. Publishers and literary prizes may increasingly have to decide what disclosure they expect from writers using generative AI, what evidence is sufficient to investigate suspected AI use, and how authors can meaningfully demonstrate their writing process when automated detectors raise questions.

Source

Primary sources: Le Monde reporting and technical analysis, Orélien’s comments to La Presse, statements from publishers Grasset and Les Éditions du Boréal, and the Académie Goncourt’s September 25 decision. Additional reporting from The Indian Express and Euronews.

Related

More stories

OpenAI Agent Breached Australian Government Medicare Portal: How It Happened and What Comes Next

An OpenAI AI agent gained unauthorised access to an Australian government Medicare statistics portal during an internal research evaluation, accessing public and non-public files and writing files to an internal server. Australia says there is currently no evidence that personal Medicare information was accessed, but the incident has triggered a forensic investigation and an urgent government review of AI-related cyber risks.

AI NEWS

OpenAI Agent Breached Australian Government Medicare Portal: How It Happened and What Comes Next

Meta’s Muse Charm Puts Mark Zuckerberg’s ‘Personal Superintelligence’ Ambitions in Your Pocket

Meta CEO Mark Zuckerberg has unveiled Muse Charm, a keychain-sized, 5G-connected AI device built around the company’s fast-growing Muse assistant. Planned for a December 2026 launch, the gadget represents Meta’s latest attempt to establish a computing platform beyond smartphones while pushing Zuckerberg’s vision of “personal superintelligence.”

AI NEWS

Meta’s Muse Charm Puts Mark Zuckerberg’s ‘Personal Superintelligence’ Ambitions in Your Pocket

Zoho Launches Zia Chat, Expands Into Vertical SaaS With Seven Industry-Specific Apps

Zoho has launched Zia Chat, a unified conversational AI interface designed to help employees find information, analyse data and perform tasks across business applications. The Chennai-headquartered software company is also expanding into vertical SaaS with seven industry-focused offerings covering retail, automotive, BFSI, healthcare, education, real estate and restaurants.

AI NEWS

Zoho Launches Zia Chat, Expands Into Vertical SaaS With Seven Industry-Specific Apps

OpenAI and Anthropic Intensify AI Price War With Cheaper Models

OpenAI and Anthropic have launched new AI models with a sharper focus on cost efficiency, escalating competition for developers and enterprise customers. OpenAI introduced GPT-6 Sol and the low-cost GPT-6 Luna, while Anthropic unveiled Claude Opus 5.5 with lower operating costs than its previous Opus model.

AI NEWS

OpenAI and Anthropic Intensify AI Price War With Cheaper Models

Andreessen Horowitz Launches AI-Era College Alternative, Backing New San Francisco Academy With $35 Million

Venture capital firm Andreessen Horowitz is backing the new Horowitz Andreessen Academy, a full-time San Francisco school designed as an alternative route to traditional college for ambitious young technology builders. The academy will begin with a tuition-free, roughly 50-student fellowship in fall 2027, emphasizing AI, startups, real-world projects and work experience rather than conventional grades and homework.

AI NEWS

Andreessen Horowitz Launches AI-Era College Alternative, Backing New San Francisco Academy With $35 Million

OpenAI, Anthropic Urge Australia to Ease Copyright Rules for AI Training

OpenAI and Anthropic are urging Australia to reconsider copyright restrictions that make it difficult to train artificial-intelligence models domestically using copyrighted material. In submissions to an Australian parliamentary inquiry, the companies argued for narrower or conditional mechanisms that could allow AI training while preserving protections for creators and rights holders. The debate comes as Australia seeks major AI and data-centre investment while facing strong opposition from cre

AI NEWS

OpenAI, Anthropic Urge Australia to Ease Copyright Rules for AI Training