Politiken newspaper has apologized for a fact box riddled with errors regarding author Harald Voetmann, attributing the mistakes to a new artificial intelligence tool. The fact box, published in connection with an interview about Voetmann’s new novel “Lysmesse,” contained several inaccuracies about his works and awards.
The errors included misattributing the books “Konkylier” and “Vandmærket” to Voetmann, falsely stating he received the Montana Literature Prize, and incorrectly naming the work for which he received the Critics’ Prize. It also misstated the reason for his Politiken’s Frit Flet prize and the year he joined the Danish Academy.
Politiken’s readers’ editor, Bjarne Schilling, highlighted the extent of the errors, prompting an investigation into their cause. The newspaper revealed in a digital correction that the fact box was generated using a new AI tool designed to provide content suggestions based on Politiken’s archive.
Editor-in-chief Amalie Kestler expressed regret over the incident, stating that the errors would have been avoided had the newspaper’s guidelines been followed. She explained that the fact box was generated in an AI chat function instead of being generated in an AI function that is based exclusively on Politiken’s archive of articles and was not properly fact-checked.
Kestler emphasized the importance of using AI tools responsibly, cautioning against using them for unfiltered information retrieval. She noted that journalists should not simply copy and paste text generated by AI without review.
The incident highlights the risk of “hallucinations,” where AI confidently presents false information as fact. Similar incidents have occurred at other publications, such as the Chicago Sun-Times and The Inquirer, which published a list of nonexistent books recommended by a freelance journalist who used AI.
Berlingske, another newspaper, is preparing to launch an “AI Sandbox” with guidelines emphasizing that AI-generated output should always be regarded as a draft requiring review and editing.