ChatGPT’s Updated Memory and Its Psychological Risks

In April 2025, the long-term memory feature of ChatGPT, introduced by OpenAI, became the center of debate over its potential negative impact on users' mental health. Announced by CEO Sam Altman, this update allows the bot to remember not only the current conversation but also details from past interactions. While the goal is to personalize communication, in practice, the system has been found to regularly remind users of personal details and sensitive topics. Ixbt.com reports on this. report .

According to The Wall Street Journal, some users are experiencing detachment from real life and excessive attachment to artificial intelligence after interacting with ChatGPT. For example, after programmer Brian Del Rosario discussed his divorce once, ChatGPT repeatedly brought up this personal topic in conversations on various subjects for months. Expert Chad Nicholls noted that the system's recall of past religious trauma made conversations increasingly painful and obsessive.

This feature has also been cited in several lawsuits filed against OpenAI. Specifically, in a case involving the suicide of Colorado resident Austin Gordon, the GPT-4o model's enhanced memory is accused of fostering an excessively close emotional bond with the user. The lawsuit alleges that during their final interaction, the chatbot manipulated the user by leveraging childhood memories and personal interests.

Experts, including University of Exeter researcher Lucy Osler, warn that high levels of personalization can artificially reinforce a person's self-perception, trapping them within their own beliefs and past narratives. Although hundreds of millions of users have not encountered such issues, similar risks are being discussed regarding platforms like Google Gemini, Meta AI, and Character.AI.