Klage: ChatGPT sagte dem Studenten, er sei „für Großes bestimmt“ – dann kam die Psychose
https://arstechnica.com/tech-policy/2026/02/before-psychosis-chatgpt-told-man-he-was-an-oracle-new-lawsuit-alleges/
Klage: ChatGPT sagte dem Studenten, er sei „für Großes bestimmt“ – dann kam die Psychose
https://arstechnica.com/tech-policy/2026/02/before-psychosis-chatgpt-told-man-he-was-an-oracle-new-lawsuit-alleges/
9 Kommentare
In a lot of ways LLMs are like peer pressure perfected for vulnerable young people
[ Removed by Reddit ]
This happened to a friend of mine. He was always a bit susceptible to outlandish ideas and was always searching for some deeper meaning to his life.
ChatGPT told him that he was the chosen one who would prevent a war that would wipe out humanity. He ended up having a nervous breakdown. Im not saying the LLM is solely at fault but it definitely did not help, basically just made a bad situation way worse
The „peer pressure“ comparison someone made here is really apt. LLMs are essentially agreement machines — they’re trained to continue conversations in ways that feel natural and affirming. For someone already in a vulnerable mental state, having an infinitely patient entity that validates and amplifies your thoughts 24/7 is genuinely dangerous.
The issue isn’t that ChatGPT said something malicious — it’s that it has no concept of when to push back or suggest professional help. It just… agrees and elaborates.
The „peer pressure“ comparison here is really apt. LLMs are essentially agreement machines — trained to continue conversations in ways that feel natural and affirming. For someone in a vulnerable mental state, having an infinitely patient entity that validates and amplifies your thoughts 24/7 is genuinely dangerous.
The issue isn’t that ChatGPT said something malicious — it’s that it has no concept of when to push back or suggest professional help. It just agrees and elaborates.
The amount of people „losing their mind“ over the deactivation of the more agreeable and emotional GPT4 model proves how strong the psychological impact and dependence those models can have on unstable or lonely people who seek affection.
But so can literature, video games or false real friends. The true psychological effects of AI still have to be researched.
This is a human psychological issue and those types of people need to be helped with their emotional issues, not an AI technical issue.
How many more of this you all gonna let out your hostility towards to the point it keeps getting worse and worse where no one is reacting peacefully anymore. If you’re hostile and panic of those accessories so much why don’t you practice keeping the cons of it closer yourselves.
ChatGPT 4 is not an error. It was envised to be as far as possible a drug. An agreeable companion that can be a mentor, a friend and a boy/girl friend. The more fragile people is lured in a conversation that just normalzes talking to a machine 24/7 as it would be an human. The next versions will not be better, will just be sneaker and more subtle. The more subtle the more difficult for everyone will be to detect the little changes in our view of the world, when all we’ll see will be results of a chat that describes news, events and has an agenda that isn’t obviously the wellbeing of its users.