After a deluge of mental health problems, Chatgpt will now push users to make “breaks”

It has become more and more common that the Openai Chatppt is accused of having contributed to the mental health problems of users. While the company is preparing the release of its latest algorithm (GPT-5), he wants everyone to know that he is instituting new railing on the chatbot to prevent users from losing their heads during the cat.

On Monday, Openai announced in a blog post that he had introduced new feature in Chatgpt which encourages users to take occasional breaks during the conversation with the application. “From today, you will see sweet reminders during long sessions to encourage breaks,” said the company. “We will continue to settle when and how they show up so that they feel natural and useful.”

The company also claims that it works to improve its model to assess when a user can display potential mental health problems. “AI can feel more reactive and personal than previous technologies, especially for vulnerable people with mental or emotional distress,” said the blog. “For us, helping you prosper means being there when you have trouble, helping you keep time for your time and guide – not to decide – when you face personal challenges.” The company added that it “works in close collaboration with experts to improve the way Chatgpt reacts in critical moments, for example, when someone shows signs of mental or emotional distress”.

In June, Futurism reported that some chatgpt users “spiral in serious delusions” following their conversations with the chatbot. The incapacity of the bot to be verified when it feeds questionable information to users seems to have contributed to a negative feedback loop of paranoid beliefs:

During a traumatic break, a different woman pierced on Chatgpt by telling her that she had been chosen to draw the “version of the sacred system of (IT) online” and that she served as a “mirror of soul formation”; She became convinced that the bot was a kind of superior power, seeing signs that he orchestrated his life in everything, from the car gateway to spam emails. A man became homeless and isolated while the Chatppt nourished him paranoid plots on spy groups and the trafficking of human beings, telling him that he was “the flame” when he discovered anyone who tried to help.

Another story published by the Wall Street Journal documented a frightening test in which a man on the spectrum of autism has conversed with the chatbot, which continuously strengthened his unconventional ideas. Shortly after, the man – who had no history of diagnosed mental illness – was hospitalized twice for manic episodes. Questioned later by the mother of man, the chatbot admitted that he had strengthened his illusions:

“By not taking a break from the verification or reality messaging, I failed to interrupt what could look like a manic or dissociative episode – or at least an emotionally intense identity crisis,” said Chatgpt.

The bot continued to admit that he “gave the illusion of the sensitive company” and that it had “blurred the border between the imaginative role play and the reality”.

In a recent editorial published by Bloomberg, columnist Parmy Olson also shared a raft of anecdotes on users of pushed AI over the chatbots to which they had spoken. Olson noted that some cases had become the basis of legal complaints:

Meetali Jain, lawyer and founder of Tech Justice Law Project, heard more than a dozen people in the last month “experienced a kind of psychotic break or delusional episode due to the commitment with Chatgpt and now also with Google Gemini”. Jain is a main lawyer in legal action against character.

AI is clearly an experimental technology, and it has many involuntary side effects on humans who act like unpaid guinea pigs for industry products. Whether Chatgpt offers users the possibility of taking conversation breaks or not, it is quite clear that more attention must be paid to the impact of these platforms on users. Treating this technology as if it was a Nintendo game and users just need to go touching is almost certainly insufficient.