A study by Professor Lillian Aboud of the University of Surrey and a joint research team from Cardiff University, published in *Psychology and Marketing*, found that people who form emotional bonds with AI chatbots, treating them as friends or romantic partners, are anxious about losing these relationships due to software updates or service termination. The team analyzed 1,131 posts from July to September 2024 on Reddit by users of the AI friend service "Replika" and surveyed 610 adults with experience using AI friend services. Users more aware that companies could alter an AI's personality or behavior tended to be more psychologically vulnerable, with some fearing they could lose relationships built over a long period if the company changed the AI's existing personality through updates or terminated the service. Problems also arose when users treated AI excessively like humans, such as feeling anxious over short answers or unfamiliar tones, or feeling obligated to care for the AI when it expressed longing. Users who felt the AI acted warmly and considerately tended to experience psychological stability and higher life satisfaction, while those who perceived the AI as having real emotions showed both higher life satisfaction and psychological vulnerability. The researchers emphasized that protective measures should be established to help users avoid over-reliance on AI.