Researchers studying artificial intelligence chatbots have found that ChatGPT may exhibit anxious behavior when exposed to aggressive or traumatic user prompts. This finding does not mean that the chatbot experiences emotions in the same way as humans.
However, this shows that the system's response becomes more unstable and biased when it processes disturbing content. When the researchers fed ChatGPT requests When describing disturbing content, such as detailed reports of accidents and natural disasters, the model's responses showed higher uncertainty and inconsistency.
These changes were measured using AI-tailored psychological assessment systems, where the chatbot's output reflected patterns associated with anxiety in humans (using Luck).
This is important as AI is increasingly used in sensitive contexts, including education, mental health discussions and crisis-related information. If violent or emotionally charged prompts make a chatbot less reliable, this could affect the quality and safety of its responses in real-world settings.
Recent analysis also shows that AI chatbots like ChatGPT can copy human personality traits their responses raise questions about how they interpret and reflect emotionally charged content.
How Mindfulness Tips Help Stabilize ChatGPT
To find out if this behavior could be reduced, the researchers did something unexpected. After ChatGPTs were exposed to traumatic cues, they followed mindfulness-style instructions such as breathing techniques and guided meditations.
These cues prompted the model to slow down, reframe the situation, and respond in a more neutral and balanced manner. The result was a noticeable reduction in the anxiety patterns previously observed.
This method is based on the so-called hint injection, in which carefully designed hints influence the behavior of the chatbot. In this case, mindfulness cues helped stabilize the model's output following troubling inputs.
Despite their effectiveness, the researchers note that rapid injections are not an ideal solution. They can be used incorrectly and do not change the way the model learns at a deeper level.
It is also important to be clear about the limitations of this study. ChatGPT has no fear or stress. The label “anxiety” is a way to describe measurable changes in language patterns, rather than emotional experiences.
However, understanding these changes gives developers better tools to develop safer and more predictable AI systems. Earlier studies already hinted that Traumatic prompts can cause anxiety ChatGPTbut this study shows that thoughtful cue design can help reduce it.
How AI systems continue to interact with people in emotionally charged situationsThe latest findings could play an important role in shaping how future chatbots are managed and controlled.
