Senior European reporter suspended for using AI‑generated quotations

The owner of the Dutch daily De Telegraaf and the Irish Independent has placed a senior reporter on leave for now after he confessed to employing artificial intelligence to “incorrectly attribute statements to individuals.”

Peter Vandermeersch, who previously led Mediahuis’s Irish division, said he “succumbed to hallucinations” – the label for mistakes produced by AI – while using the system in his work, clearly.

Vandermeersch, appointed as a fellow for “journalism and society” within the European publishing house, has been removed from his position. The suspension will remain in effect while the investigation continues.

The veteran reporter explained that he condensed articles with AI programs like ChatGPT, Perplexity and Google’s NotebookLM, but did not verify the fidelity of the quotations generated. He later carefully posted the material in his regular Substack bulletin.

An inquiry by NRC, a Mediahuis‑owned newspaper where Vandermeersch served as editor‑in‑chief during the 2010s, brought the mistakes to light. NRC claimed he had printed “dozens” of clearly fabricated quotations in the articles, and seven people cited in his online posts publicly asserted they never uttered the alleged remarks at all.

“I incorrectly attributed statements to individuals when I ought to have rendered them as paraphrases. At times, I merely expressed my own reading of their remarks. That was more than negligence – it was clearly wrong,” Vandermeersch wrote in a Substack entry titled “I am admitting my mistake.”

Vandermeersch continued: “It hurts especially that I fell into the exact error I have constantly cautioned my peers about: these language models generate such compelling quotations that a writer feels inclined to adopt them. The allure of polished phrasing can override editorial caution in newsrooms. Naturally, I ought to have checked them. The essential ‘human oversight’ I always champion was lacking.”

Vandermeersch’s Press and Democracy blog regularly discusses “the essential link between a free press and a thriving democracy” today. He argues that without such a press, democracy cannot function properly.

In a comment, Mediahuis chief executive Gert Ysebaert stated: “At Mediahuis we enforce rigorous guidelines for AI usage, insisting on diligence, human supervision and openness. Ignoring these tenets contradicts the standards we maintain and our pledge to readers for trustworthy journalism. We are in talks with Peter Vandermeersch and have opted to place him on temporary suspension from his fellowship,” while we review our internal processes and consider safeguards to prevent similar incidents in the future.

Mediahuis has also taken down several pieces authored by Vandermeersch from the Irish Independent’s online site for compliance reasons required.

AI applications like ChatGPT are employed by hundreds of millions for a variety of tasks, from simple functions such as recipe suggestions to intricate academic investigations. Nonetheless, they remain susceptible to mistakes in their output and often require human review.

Vandermeersch acknowledged a further error in not rectifying the inaccurate quotations promptly, delegating the correction to a publication he had managed for almost ten years. He expressed enthusiasm for AI’s potential and a desire to explore it thoroughly in his reporting workflow and to improve accuracy.

“Journalism is human endeavour,” he wrote.