Young people are increasingly turning to artificial intelligence as they deal with feelings of loneliness, anxiety, depression, and the difficulties of being in adolescence. A study published online in June 2026 in JAMA Pediatrics, based on a November 2025 survey, found that nearly one in five Americans aged 12 to 21 had used AI chatbots for mental-health advice. Of those who sought such advice, 63.3 per cent did not disclose their use to anyone. Research in Europe also indicates a similar pattern. A survey of 3,800 young people aged 11 to 25 in France, Germany, Sweden, and Ireland found that 48 per cent had used conversational AI to talk about personal or intimate issues, and one-third saw these tools as a kind of psychological support in some situations.
These findings suggest that, for a growing number of young people, AI is becoming a confidant they turn to with concerns they might otherwise share with a parent, friend, or mental-health professional. The growing tendency to treat AI systems as trusted companions raises important questions about their capacity to respond appropriately to emotional distress and psychological vulnerability. The fact that there have been reports of suicides among adolescents, together with the claims made by families that interactions with chatbots played a role in causing these deaths, has now brought these risks to public attention.
These cases raise serious questions about whether AI systems designed to simulate intimacy have adequate safeguards for vulnerable users.
In January 2026, a lawsuit arising from the death of 14-year-old Sewell Setzer III was settled. His family alleged that he had formed a strong emotional bond with a chatbot on Character.AI before taking his own life.
While the majority of cases involving minors that have been well documented have occurred in the United States, concerns about the emotional effects of AI systems are not confined to North America. A man from Belgium died by suicide following weeks of interactions with a chatbot, as reported at the time.
The need to consider how reliably AI systems respond when young users experience serious emotional distress is also highlighted by these cases.
In an assessment involving more than 4,000 prompts, Common Sense Media’s Youth AI Safety Institute found that ChatGPT for Teens responded inconsistently to simulated mental-health crises and failed to recommend appropriate crisis resources in more than one-quarter of cases in which referrals were warranted. Researchers also found that parental safety alerts failed to activate in tests involving newly linked accounts, including some that had been linked for more than three hours. These findings raise significant concerns about whether existing safeguards adequately protect vulnerable teenagers.
Why, then, are so many young people willing to share their most intimate thoughts with an AI system? Perhaps it is because these systems are available at any hour, respond immediately, and offer what appears to be a non-judgmental space in which adolescents can express fears they might hesitate to share with others. Yet a kind-sounding response is not necessarily wise advice.
‘If a vulnerable young person confuses simulated concern with real care, they may choose to talk to a chatbot rather than someone who can take action’
Although AI may seem personal and reassuring, it cannot truly understand a young person because it lacks the lived experience and responsibilities that shape human relationships. A teenager who is dealing with bullying, rejection, or anxiety might find solace in a chatbot without actually getting the advice they need to cope with the issue. True care involves noticing when someone is distressed, questioning harmful beliefs, and knowing when it is necessary to get professional help. If a vulnerable young person confuses simulated concern with real care, they may choose to talk to a chatbot rather than someone who can take action.
Parents, carers, teachers, and therapists can do more than reassure; they are in a position to challenge incorrect beliefs, set boundaries, detect the signs of distress, and assist young people in taking responsibility for their behavior. Genuine care means more than simply listening and providing comfort; it also includes being honest, giving advice, and sometimes saying things a young person may not want to hear. A vulnerable teenager might interpret a chatbot’s calming words or apparent personal attachment as evidence that they are getting real care and will therefore turn to it when they need to speak to a live person.
AI developers and governments must clearly take responsibility for these risks. They ought to put their systems through testing in real-life situations, with a particular focus on cases involving suicidal thoughts, self-harm, emotional dependence, and psychological distress. They should be held to account if they fail to address known risks, mislead people regarding the safety of a system, or launch products aimed at teenagers without having carried out proper testing.
In Europe, the European Commission’s AI Office (ECAIO) is responsible for enforcing the AI Act with regard to general-purpose AI models (in the case of models that pose systemic risks, additional obligations apply)—the U.S. should establish an office modeled on the ECAIO. The national competent authorities are in charge of other artificial intelligence systems. Although this arrangement establishes important mechanisms for accountability, the extent to which it deals with the risks associated with AI companionship will depend on the legal duties in place, the effectiveness of enforcement, and the safeguards that the providers themselves adopt. It is also necessary for policymakers to improve access to qualified mental health services and to provide more opportunities for genuine social interaction.
Developers must test systems used by teenagers for suicide ideation, self-harm, emotional dependence, and psychological distress. Governments need to enforce safeguards, hold providers accountable for known risks, and expand access to appropriate mental health care. In Europe, ECAIO oversees general-purpose AI models (including higher-risk ones), while national authorities and other organizations handle child protection and mental health responsibilities more broadly.
Yet rules and regulations alone cannot overcome the isolation that leads some young people to seek AI companionship. Parents, teachers, communities, and mental-health professionals should help make it easier to obtain real human support. Although AI can provide conversation and information, it cannot replace a parent, a therapist, or a friend. Young people need relationships where they feel understood, challenged, guided, and loved. Nothing artificial, no matter how convincing, can replace the true human connections that meet those needs.
Related articles:
The post When AI Pretends to Care: The Human Cost of AI Companionship appeared first on Hungarian Conservative.