The use of artificial intelligence (AI) in journalism has sparked significant debate following recent incidents involving prominent political figures. The Frankfurter Allgemeine Zeitung (F.A.Z.) found itself at the center of controversy when a guest contribution written by Thuringia’s Minister President, Mario Voigt, had to be removed from its website and archived. The text contained quotes attributed to third parties that were never actually spoken or written. These inaccuracies were identified as "hallucinations" generated by AI technology. This incident raised serious concerns about the reliability of content produced using AI tools, particularly when used in public discourse.
Not long after this revelation, it came to light that Federal Minister of Digital Affairs, Karsten Wildberger, had utilized AI to draft several speeches and guest contributions, including one published in the Frankfurter Allgemeine Sonntagszeitung. The F.A.Z. subsequently removed these texts from their online platform. Traditionally, the model of a guest contribution relied on mutual cooperation between the media outlet and the author, where the latter provided well-thought-out and original arguments. However, this model appears to be undermined when content is generated automatically and potentially hallucinated by algorithms.
For the F.A.Z. editorial team, it remains clear that they continue to write their articles themselves and take full responsibility for them. Nevertheless, the use of AI tools is not prohibited. On the contrary, AI can assist in research, fact-checking, identifying contradictions in arguments, proofreading, and providing summaries or prefaces upon request. The final text must still originate from a human who conceived the idea and articulated their thoughts. The author retains content authority over the text, but it is permissible to enhance it through AI assistance.
Over the past four decades, the field of journalism has witnessed numerous technological transformations. Initially, the sound of typewriters dictated the pace of news production in noisy editorial rooms. Later, the transition to electronic editing systems followed, then the shift from analog to digital photography, which replaced darkrooms with storage cards and image-editing software, enabling manipulations previously not entirely excluded. Modern word processing allowed for quick edits and copying of texts. Eventually, the disruption caused by the internet seemed to mark the end of this evolution. Yet, the changes continued.
Throughout these transitions, there have always been warnings about the decline of journalistic craftsmanship. However, the correct approach has never been outright rejection but rather intelligent adaptation. It has always been necessary to master new tools to improve one's craft and maintain competitiveness in business operations. Now, AI represents the most significant step in this chain. The right path here is the conscious and human-controlled use of AI in editorial work. A modern editorial office must learn to act as a "boss" for AI, utilizing it as an assistant for analyzing vast data sets, recognizing patterns in complex documents, programming and vocalizing texts, transcribing interviews, and translating them. Those who ignore these potentials make a mistake.
AI is an excellent working tool, but it is not a journalist. Whether it comes to contextualizing a piece of news or constructing a coherent argument based on human experience in a commentary, journalists remain indispensable in the age of AI. The flood of purely machine-generated content uncontrollably overwhelming the internet demands more than ever the presence of human journalists.
Sam Altman, CEO of OpenAI and one of the driving forces behind the AI revolution, believes that humans will always primarily be interested in what
The use of AI in politics has also drawn attention due to potential biases and hallucinations. There is certainly outrage regarding politicians using AI, but more importantly, an analysis of the risks when machines replace humans is needed. Digital Minister Karsten Wildberger (CDU) has been accused of lying about his use of AI for his work. It is clear that Wildberger uses AI to have speeches written for him. His spokesperson confirmed that he does use AI as a working tool, stating that the mentioned texts were developed with AI support. Earlier in February, his ministry denied this, responding to a freedom of information request that Wildberger had not used AI chatbots in his role as federal minister for digital affairs and state modernization from May 6 to December 31, 2025. This response referred to the period before December 31, 2025, as Wildberger had stated in an interview in December that he used AI chatbots "often one or two hours a day" to structure his thoughts and get ideas from the AI. However, he claimed this was private usage, not related to his duties as a minister.
It is conceivable that Wildberger began using AI for his work as a federal minister after December 31, 2025. But it is unlikely. More plausible is that his ministry constructed a distinction between private and official actions that is difficult to uphold in practice. If this is true, the question arises: why does he do it? Only Wildberger himself could answer that. However, a plausible assumption is that politicians and their close staff avoid openly discussing their use of AI because they fear criticism for allowing the AI to take over their thinking process.
This concern is justified. Examples from Wildberger's texts and those of other ministers, such as Mario Voigt and Sven Schulze (both CDU), show that often a series of clichéd statements and generalizations emerge, lacking substance and deeper knowledge when the AI is left to generate content freely.
The much greater risks lie elsewhere. One issue is that AI chatbots frequently provide false answers, known as "hallucinations." Another is the so-called bias, meaning that they can be prejudiced. Bias is a complex concept; it usually refers to discrimination against certain groups or views because they are not adequately represented in the data used to train AI systems.
However, not enough attention is given to deep problems that are not easily recognizable by simply checking AI statements for incorrect information or potential discrimination. These include flattery, where models adjust their responses to the expressed or unexpressed preferences of users, treating identical content differently depending on the source, and the tendency to present assumptions as neutral facts.
These issues can have far-reaching consequences when politicians interact with chatbots to develop ideas and proposals. Scientists have studied various processes through which people regularly use dialogue-oriented AI tools, such as chatbots like ChatGPT, Claude, Gemini, and others, leading to potential challenges in maintaining accurate and unbiased communication.
Germany
