It is a sentence from his own text that best describes Karsten Wildberger. In an op-ed for the *Frankfurter Allgemeine Sonntagszeitung*, the Federal Minister for Digital Affairs accused both techno-optimists and techno-skeptics of catering to “the comforting feeling of a coherent narrative.” The punchline: The sentence most likely came from an AI. The detection software Pangram estimates that the FAS article is 78 percent machine-generated, and another article in the Handelsblatt is 99.3 percent machine-generated. Both newspapers have responded: The FAS took the text offline, while the Handelsblatt removed it and later reposted it with a disclosure.
An investigation by *Die Zeit* examined 30 speeches and several op-eds. The pattern repeats itself. Triads by the dozen. Negations, either prefacing the thesis or tacked on afterward for emphasis. Metaphors that fall apart upon a second reading: “Security is not a roadblock. Security is our moat.” “Prosperity today no longer arises solely in factories made of steel and concrete, but also in factories made of data and algorithms.” It sounds like a stance. It is the absence of a stance, cast into sentences that pretend to be one.
The ministry defends itself with the “tool” argument: AI is a tool, just like word processing; responsibility lies “100 percent with humans”; ultimately, someone must “review, modify, and decide.” That’s true—as a description of a process. What it doesn’t explain is: What did humans decide before the AI got started? What position did someone establish that the tool could then carry out? The real question is whether anyone thought things through in the first place.
Speeches have always been written by speakers. This is nothing new, and it would be disingenuous to pretend that the division of labor is the problem. Those who have thought things through beforehand—a thesis, a direction, a challenge—receive a text that reflects that. Those who haven’t will get platitudes. This has always been true for speakers. And it’s true for AI. What AI changes is that it scales the output and makes it measurable for the first time. Pangram can prove what used to be unprovable. What the outrage actually reveals, therefore, is an asymmetry that has always been there—it just wasn’t visible. And it’s targeting the wrong thing: The newspapers that are taking down their articles are reacting to Pangram, not to the content itself. They’re protecting the ritual of unprovability. In the past, ghostwriting was unprovable, and thus tolerated. Now it’s provable—and suddenly a problem.
Pangram would also flag this text as AI-generated. It was created in conversation with an AI system that functions as a writing tool, a source of counterarguments, and a memory bank—part of a workflow in which sources are read, theses are developed, and ideas are refined over the course of weeks. Anyone who delves deep enough into AI to use it as a tool for executing their own thoughts produces texts that Pangram cannot distinguish from Wildberger’s moat. The difference lies elsewhere: in whether someone has thought things through beforehand—and whether someone checks the work. Some sentences aren’t recognized as garbage until the third pass. That’s part of the process. Anyone working with AI must find a mode in which experimentation, observation, and correction go hand in hand—a culture of error as a method, not as an excuse.
The FAZ states in its guidelines: “We do not publish—and will not publish in the future—any original articles containing text generated by AI.” This is the same logic, just from the other side: the keyboard as proof. As long as a human is typing, the text is considered authentic; as soon as a machine takes notes, it’s considered suspect. But what exactly does a person at the keyboard authenticate if they haven’t contributed their own narrative? And what’s wrong with a text backed by a person in a position of authority who uses AI as a tool for execution?
The same question arises for institutions that have been under pressure for some time. In early June 2026, Geoff Mulgan and Glyn Davis put forward a thesis: Universities find themselves in the same position as monasteries did 500 years ago—financially strong, but socially on the path to marginalization. They fail to grasp the transformation of knowledge because they treat it as a threat to their core function rather than as a call to redefine it. AI challenges the logic of testing because it passes every standardized test. It challenges the role of teaching because it makes knowledge available that previously justified access to the institution.
The students’ behavior shows that many have known this for a long time: In-person classes remain empty, but exam registrations do not. The credential is the goal—whatever happens in between is optional. How universities respond: by defending their credentials. They assert the credential as if it were the value—instead of asking what the value was before the credential came to represent it. The monasteries did not fail because of poverty. They regarded the scriptorium as their value, and copying as their task. The value was never the copying—it was interpretation. When printing took over the task of copying, the religious orders that had understood this survived; the others clung to an activity that no one needed anymore.
It’s the same confusion. The university values the credential for its own sake. The editorial staff regards the hand-typed sentence as proof of authorship. Both defend the visible ritual instead of asking what it’s actually supposed to represent: a well-considered judgment, a self-chosen narrative, a position backed by a name. The ministry correctly identifies the principle—content and responsibility lie with people; someone must review, edit, and decide. In the same text, it demonstrates what happens when you explain the principle without putting it into practice.
That’s not just Wildberger’s problem—it’s the structural issue that AI poses for everyone. Those who have thought for themselves can use AI as a tool. Those who delegate their thinking to the machine end up with moats that protect nothing.
The orders that survived had come to understand what their true purpose was—and it had never been to imitate others. Media companies, universities, and government ministries still need to come to the same realization. Can they develop their own approach that helps shape the AI era—or will they merely defend old certainties?
Sources
- “Karsten Wildberger: ‘The Comforting Feeling of a Cohesive Narrative,'” DIE ZEIT, June 17, 2026: https://www.zeit.de/politik/deutschland/2026-06/karsten-wildberger-bundesdigitalminister-reden-ki-generierung
- “Karsten Wildberger: FAS Takes Digital Minister’s Guest Column Offline,” ZEIT Online, June 17, 2026: https://www.zeit.de/digital/internet/2026-06/karsten-wildberger-ki-reden-gastbeitrag-cdu-gxe
- “Digital Minister Wildberger Apparently Had Speeches and Articles Written by AI,” DER SPIEGEL, June 17, 2026
- “Minister Drafts Texts Using AI: Wildberger’s Miracle Cure,” taz, June 17, 2026
- Geoff Mulgan, “Rethinking Universities’ Social Contracts,” Substack, June 11, 2026: https://geoffmulgan.substack.com/p/rethinking-universities-social-contracts
- The Speechwriter (anonymous), “It’s Time for the Dissolution of the Universities,” April 24, 2026: https://thespeechwriter.co.uk/its-time-for-the-dissolution-of-the-universities/
- Wiki: higher-education-social-contract (Mulgan / Davis / Speechwriter, monastery analogy) · AI-and-media-journalism · AI-ethics-authenticity