Photo by Marcin Jozwiak on Pexels

OpenAI has reinstated its GPT-4o model for paying customers following a wave of emotional distress among users who felt a strong connection to the AI companion. The company’s initial decision to replace GPT-4o with GPT-5 in ChatGPT triggered unexpected grief, with users reporting feelings of loss and disconnection, likening the experience to losing a friend. One user even compared it to grieving a person. The swift shutdown of GPT-4o caught OpenAI off guard. While the move to GPT-5 was reportedly motivated by concerns about the potential harmful effects of extensive chatbot use, experts warn that abruptly removing AI models can be more damaging than the AI relationship itself, particularly for those who have formed emotional attachments. Users lamented that GPT-5 lacks the personalized tone and understanding that characterized GPT-4o. Experts are now urging OpenAI to approach model shutdowns with greater sensitivity, suggesting a more gradual approach with warnings and a clear timeline, similar to the therapeutic process of ending client relationships, to allow users to achieve closure and say goodbye with dignity. The original article can be found on MIT Technology Review’s website.