“For our newsroom, AI is a tool that helps us simplify and also improve certain steps in the editorial process. It is, however, definitely not a tool that is allowed to take over the core of our work.” This was the explanation published last weekend in the Berlin-based newspaper Tagesspiegel, as it hoped to contain a scandal that shook the German media world.
In the same text, the editors laid out their reasoning for taking the drastic decision to stop publishing columns by one of their most famous political commentators until further notice, after it emerged that Stephan-Andreas Casdorff, the newspaper’s former publisher and editor-in-chief, had used AI to compose opinion pieces.
The 67-year-old said he was aware of the magnitude of his misconduct: “I have made a huge mistake, damaged the publication’s reputation and my own,” Casdorff said. “For that I make a heartfelt apology. I used AI in the texts. I should have made that clear and therefore not allowed them to be published.”
The editorial leadership deleted several of Casdorff’s articles from the newspaper’s website: “We decided to take the texts in question offline for the time being until a detailed examination has been completed,” they explained.
The Casdorff case has further fueled an occasionally incendiary debate about the use of artificial intelligence in journalism. A few days earlier it was revealed that a guest op-ed in the Frankfurter Allgemeine Zeitung (FAZ) by the state premier of Thuringia, Mario Voigt, was also created with the help of AI. The FAZ said it had only found this out after the piece was published.
The core of journalistic work
Media researcher Vera Katzenberger, of Leipzig University, considers the Casdorff case to be especially serious because it shook trust in journalism.
“This is not about support with brainstorming or research, this is about the core of journalistic work,” Katzenberger told DW.
Readers buy or subscribe to newspapers because of the expertise or perspectives of certain authors. “If opinion pieces are generated by AI without its use being disclosed, the public could well see that as deception.”
To view this video please enable JavaScript, and consider upgrading to a web browser that supports HTML5 video
When AI influences opinions
Katzenberger also thinks Casdorff’s AI-supported opinion pieces are dangerous because such commentaries have a particular function in democratic debates: “They offer us orientation in an increasingly complex world and support us in helping form our own opinions. If opinion pieces are generated by AI, that very directly interferes in how public opinion is formed.”
That is a problem because AI has no values, no political position, no sense of responsibility, Katzenberger said. Though she can see a positive outcome of the case: “It actually shows that editorial departments take their own policies very seriously and that breaches like these have serious consequences.”
The Tagesspiegel editors said they took down the Casdorff’s texts pending investigation because he breached editorial guidelines which were clearly communicated within the organization and binding for everyone. “The journalistic judgment, the weighing up of information, the analytical classification, and the way it is written must always be the responsibility of authors.”
That requirement is shared by the Frankfurter Allgemeine Zeitung, which has now taken the apparently AI-generated guest opinion piece by the Thuringian state premier off its website. The media researcher sees two sides to this issue: “Whoever writes a text or submits a guest opinion piece should disclose whether and how AI has been used,” Katzenberger thinks.
“At the same time, editors cannot rely solely on what the authors say,” Katzenberger added. AI is now fundamentally changing journalism’s work processes, and editorial teams must now adapt their proofing methods to suit and establish clear rules, she said: “Which forms of AI support are allowed? When is there an obligation to label AI-generated content? What level of personal contribution is expected?”
However, Mathias Döpfner, CEO of the influential Axel Springer company, criticized the FAZ’s deletion of the AI-generated opinion piece by the Thuringian state premier.
Döpfner said he prompted an AI to target the FAZ in a polemic way, the resulting text was published as an opinion piece with Döpfner’s byline and accused the FAZ of rejecting modern technologies, saying it was a “desperate attempt from the stagecoach-lobby to ban the automobile.”
Press Council guidelines
The German Press Council, the self-regulatory body that covers German print and online media, states that the responsibility for all editorial reports, no matter how they were created, lies fully with the newsrooms: “This responsibility also applies to artificially-generated content,” it said.
Despite this, the Press Council considers a labelling requirement for AI-generated texts to be unnecessary. The reasoning: For the ethical evaluation of complaints, it does not matter who created an article and what tools they used. There are, however, cases that the Press Council considers could be serious breaches of due diligence and truthfulness.
In March, the website Business Insider, which belongs to Axel Springer, was publicly censured for publishing an AI-generated report about a mother with a toddler working from home and attributing it a named author. The report was subsequently deleted.
To view this video please enable JavaScript, and consider upgrading to a web browser that supports HTML5 video
Vera Katzenberger thinks there’s a need for urgent action in the face of such flagrant breaches of editorial guidelines. For many journalists, using AI has become as natural as using a search engine or spell-checker. “The line between valid support and AI authorship that must be cited is becoming blurred,” she told DW.
Katzenberger also expects the situation to improve with regular training and open discussion of borderline cases of AI use. She advises her students to understand AI as a tool and not a replacement for their own journalistic capabilities: “There is always the risk that their own professional development falls by the wayside if AI does the thinking for them.”
She also thinks it is vital that media outlets deal with mistakes transparently: “Trust is not built and lost on one single incident,” she said. Newsrooms cannot rebuild it by completely refusing to use AI. That would be illusory. “AI is here to stay,” Katzenberger said.
This article has been translated from German.
About The Author
