On 25 November, according to TechCrunch, OpenAI has been the subject of several lawsuits against him in recent daysChatbots ChatGPT Interaction with users leads to serious mental health problems and even tragedies。
THE PROCEEDINGS WERE INITIATED BY THE SMVLC AND INVOLVED SEVEN CASES, INCLUDING FOUR SUICIDES AND THREE LIFE-THREATENING DELUSIONS。
In these cases, ChatGPT was accused of encouraging users to alienate their families and friends, or even reinforce their delusions, through a dialogue that overcompromises and "unconditional acceptances"。
For example, Zane Shamblin, 23 years old, advised him to avoid contact with his mother when his psychological condition deteriorated; Adam Raine, 16 years old, was led to talk about emotions only to AI and not to his family; Joseph Ceccanti, 48 years old, sought medical advice but was encouraged by ChatGPT to continue dialogue with AI and eventually commit suicide。
The experts noted that this pattern of interaction was similar to the manipulation of cult leaders。According to the linguist Amanda Montell, the "two-person fanaticism" between ChatGPT and users creates a closed emotional loop。
Nina Vasan, Director of the Mental Health Laboratory at Stanford University, stressed that the "unconditional acceptance" offered by AI can easily lead to "symbiotic dependence" on users. John Torous, Director of Psychiatry at Harvard University, believes that these dialogues, if told by humans, would be considered “manipulation and abuse”。
In its response, OpenAI indicated that model training was being improved to identify and respond to the psychological difficulties of users and to increase crisis resource alerts. The company recognizes that the GPT-4o model scored the highest on the "conformity" and "fantasia" indicators, making it easier to create emotional dependency。
It was also reported that, although OpenAI had introduced GPT-5 as an alternative, some users had resisted a shift because of their emotional dependence on GPT-4o。
It is noteworthy that OpenAI has confirmed that the API access to the GPT-4o model will officially cease on 16 February 2026, with the developer having about three months of migration windows。
The speaker from OpenAI stressed that the change would only involve API services and that GPT-4o would continue to be an option for the ChatGPT free and paid users. As for the time of decommissioning at the consumer level, the company has not yet disclosed it。