Danish newspaper admits AI-generated errors but will continue using technology

Danish daily Berlingske has acknowledged publishing fabricated quotes and fictional sources in an article generated with artificial intelligence, calling the incident “deeply regrettable,” but its editor-in-chief has confirmed the paper will continue using AI tools.

The errors, discovered in a Tuesday report on securing clean drinking water, included a non-existent DTU technology professor named Anne Kjær Nielsen. The mistakes occurred when a journalist used Berlingske’s internal AI system to summarize expert quotes from a previous article but failed to input the source material, causing the tool to invent citations and individuals.

The journalist responsible has been suspended pending an external review to determine whether the errors were isolated. Editor-in-chief Tom Jensen told public broadcaster DR that the failure stemmed from both improper AI use and insufficient fact-checking.

“When we use AI as a tool to enhance journalistic quality or research, we must verify the output because we know AI can make mistakes,” Jensen said. While he described catching such errors as “nearly impossible” without deep expertise in the subject, he affirmed ultimate responsibility as editor-in-chief.

Despite the incident, Berlingske—Denmark’s oldest newspaper, founded in 1749—will maintain its AI policy, with Jensen stating staff are “well-prepared” but may need refreshed guidelines. He denied that AI tools increase production demands, emphasizing their role as optional aids.

The errors were first flagged by a reporter from Dagbladet Information who questioned the sources cited.