A man in Zurich attempts to kill kindergarten children with a knife. In court, a reconstruction of how he prepared for the crime is taking place. And in the middle of this reconstruction, a term comes up that is sure to grab attention these days: ChatGPT. The perpetrator reportedly communicated extensively with ChatGPT and, among other things, asked where he could find young children in Zurich. At the same time, there is mention of internet research, the purchase of a knife, anatomical studies, and practice on meat. Read the article
So the mention of ChatGPT is relevant. But it is not yet an explanation.
The crucial information is missing
Anyone who wants to know what role the AI system actually played would need to know two things in particular: What was asked—and what did ChatGPT respond?
We know part of the first part. We don’t know the second. Did the system provide specific information? Did it refuse the request? Did it refer the user to help? As long as we don’t know that, it’s hard to deduce from mere usage what role ChatGPT actually played in the planning of the crime. Usage does not automatically equate to impact. And mention does not automatically imply causality.
Equally important is the question of whether the same information could not have been found through search engines, forums, videos, or websites. Generative AI is changing the way we access knowledge. That is precisely why we must carefully examine what it actually contributed in a specific case. “He used ChatGPT” is, in and of itself, hardly any more meaningful than “He used the Internet.”
The key journalistic question , therefore, is: What information did he receive from the AI that was relevant to the crime—and could he have found it elsewhere as well?
AI can also intervene
Another point that is often overlooked in the debate is this: Generative AI is not just a new information tool. It can also detect problematic intentions, refuse to provide answers, offer safe alternatives, or refer users to sources of assistance. In this way, it fundamentally opens up possibilities that traditional information systems do not have.
More differentiation would be helpful
If an AI system is involved in a crime, that should of course be included in the news coverage. But the next question is just as important: What actually happened in that interaction? Perhaps the AI facilitated the crime. Perhaps it rejected a problematic request. Perhaps it was just one of many tools.
AI deserves critical scrutiny. But criticism isn’t strengthened by hasty conclusions about causality—it’s strengthened by careful examination. That’s why, whenever we see a report stating, “He did his research using ChatGPT,” one question should come to mind: And what did ChatGPT say?