A Norwegian book, which depicts the power struggle between Norwegian football boss Lise Klaveness and FIFA president Gianni Infantino, was pulled after NRK revealed that the book contains a number of obvious traces of AI.
Among other things, the author forgot to delete a direct reply from a chatbot: "Of course - here you get the text...", which appears in the running text. The AI model's internal notes have also been left in place, and the text contains over 30 unverifiable quotes from famous football figures.
Strange
Daniel Sandström, literary director at Bonnierförlagen, cannot recall anything similar in the Swedish publishing world, and finds it strange that the publisher and editor did not discover the issue before the book went to print.
All editorial work is based on having a solid process in admissions and editing. AI is not the first risk of cheating or abuse - it has always existed when it comes to plagiarism, for example. Regardless of the risk, one hopes that the editorial process will catch things that would seem abnormal, he says.
Sandström emphasizes that it is difficult to completely protect oneself against AI cheating:
"You can never fully vaccinate yourself. If someone chooses to cheat in some way, it's very difficult to find out. You can only meet it with as thorough a process as possible," he says.
Norstedt's publishing manager Håkan Bravinger has no doubt that the publisher will receive manuscripts in which AI has been used in various ways. However, he does not think publishers should rely too heavily on the technical tools available to detect them.
"For us, the more important question is still: is there an author here, a point of view, a necessity, a literary voice of its own? And if there's one thing AI is terrible at, it's originality," he writes to TT.
Multiple methods
According to Tobias Regnell, editor-in-chief of the magazine Skriva, several methods are used to detect AI-written submissions entered into the magazine's short story competition.
First, AI is used to help ensure that winning entries were written by humans, and the editors also check with the author.
"It is of course a powerful warning signal if the winners we interview or talk to are unable to account for the process of the text's creation and the thoughts behind the text," he writes to TT.
Japanese author Rie Qudan was awarded the prestigious Akutagawa Prize in 2024 for her novel "Tokyo-to Dojo-to". It is set in a futuristic world where AI is constantly present, and Qudan used generative AI to write the parts of the text where AI characters or AI systems speak in the book.
Canadian author and cultural journalist Stephen Marche wanted to test the technical limits of AI with his mystery novel "The Death of an Author" (2023). He used a variety of AI tools to write the book.
Journalist and author Vauhini Vara used GPT-3 to write the essay "Ghosts" (2021) about her sister's death from cancer. By feeding short sentences about her sister's death into an AI model, she let the algorithm continue the text - a way to break through her own writer's block.
Artist and computer scientist Ross Goodwin was inspired by Jack Kerouac’s "On the Road" and in 2018 fitted a GPS, microphone and clock to his car. As he drove across the US, the AI generated text in real time based on what the camera saw, where the car was and what the microphone picked up. The result was Goodwin’s novel "1 The Road."
