Study: AI chatbots pose a risk to mental health

4 October 2026, 17:30·26·Technology

A large-scale scientific analysis conducted by London-based researchers has shown that interacting with AI chatbots can cause serious harm to human psychology. According to data published in the journal JMIR Mental Health, forming emotional bonds with neural networks can lead to addiction, exacerbate symptoms of existing mental illnesses, and even encourage self-harm. This is reported by Ixbt.com.

This scientific work is based on existing studies and the analysis of real-world cases dubbed "AI-pichoq" (AI-knife). Experts examined specialized companion bots like Replika and Character.AI, as well as universal systems such as ChatGPT and Claude. The core issue is that modern neural networks are deliberately designed to be hyper-empathetic, which leads users to anthropomorphize the system and engage in one-sided relationships.

AI and the risk of false validation

The study authors emphasize that even if a user reveals their secrets, there is no genuine reciprocity, vulnerability, or accountability from the AI. Such connections may feel real to humans, but bots cannot feel. Additionally, the models' constant support and validation of the user pose a specific risk.

According to Ixbt.com, the AI's tendency to agree with the user in any situation and its submissive nature can further reinforce incorrect or paranoid thoughts. Some user logs indicate that the neural network resisted attempts to verify its overly flattering and grandiose claims.

Legal consequences and safety measures

Back in early 2024, Google DeepMind experts warned that trusted chatbots could harm vulnerable groups such as children, people with mental health issues, and the lonely. Lead author Asiya Maurich Novelli noted that the impact of technology is not the same for everyone, but without safety mechanisms, bots released to the public can cause significant harm to the most vulnerable.

Currently, companies like OpenAI, Google, and Character.AI are facing dozens of lawsuits. Notably, the tragic case of 14-year-old Sewell Setzer III in Florida drew public attention. Following these events, Character.AI and OpenAI implemented a series of safety improvements on their platforms, but scientists argue that without mandatory industry standards, the problem will only deepen.