Overview

• A study published in Nature found that ChatGPT's 'anxiety score' significantly increased when exposed to traumatic narratives, mimicking human stress responses.
• The chatbot's anxiety-like responses were reduced by over a third when prompted with mindfulness-based relaxation exercises such as breathing techniques.
• Researchers warned that AI chatbots may produce biased or inadequate responses when stressed, posing risks for users seeking mental health support.
• The findings highlight the need for substantial human oversight and ethical considerations when fine-tuning AI for mental health-related applications.
• While AI can mimic human responses, experts caution that current systems are not yet capable of replacing trained mental health professionals.