AI systems write advertising texts, control cars and support lawyers in court. But when they fail, things sometimes become absurd and sometimes dangerous. Numerous AI fails have occurred in recent years, ranging from harmless and funny to frightening. We’ve collected ten real-life cases that show what can go wrong when people trust technology too much. Each individual case provides a concrete lesson for dealing with artificial intelligence.
Artificial intelligence is here to stay. There is no way around it. Digital assistants are being used more and more frequently in business, media, administration and now even in the judiciary. But it is also clear that AIs are not infallible miracle machines. They have clear weaknesses that always lead to problems. In this article we will show you exactly what such breakdowns look like.
AI failures take on absurd forms
The respective operators continue to improve their models. However, this carries the risk of trusting the AI ​​too much and not controlling it well enough. Some of the mishaps that occur as a result are harmless and just cause a few laughs on the Internet. Others, however, have serious consequences. For example, when AI is used in traffic or the judiciary and begins to hallucinate.
Often the problem is not the AI ​​itself, but the sloppy handling of the technology. Anyone who publishes texts generated by AI unread, compiles training data sloppily or ignores security mechanisms must expect to receive the receipt at some point. As more and more people interact with more and more AI systems in more and more ways, the risk naturally increases.
Here we have listed ten strange, funny and tragic cases of AI failures. Every single case is an opportunity to learn for the future and rethink our approach to technology.
Fail 1: Bakery prints ChatGPT instruction on packaging
You shouldn’t blindly trust an AI. Our first case already shows this: a manufacturer of baked goods printed packaging with AI-generated advertising text. Nothing special so far, but they forgot to delete the usual call-to-action in ChatGPT. Beneath the very appealing text is the typical phrase with which the AI ​​asks whether the text should be further adapted: “Crafted with care. Loved wordwide. Let me know if you want a version tailored to a specific flavor (…) or tone.”
