Building on Indigenous Knowledge for Strengthening Food Security and Nutrition: Evidence from a Community-based Intervention in Eastern India.

Adolescent mental health is increasingly recognized as a defining public health challenge of the digital era. One in 7 young people aged 10 to 19 years lives with a diagnosable mental health condition, reflecting a substantial and growing burden of disease during adolescence. Parallel increases in digital media exposure have prompted renewed scrutiny of how emerging technologies shape developmental environments. Over the past decade, social media shifted from a communication tool to a potential public health exposure, with ongoing debate regarding its association with adolescent well-being. Population-level analyses have also highlighted broader shifts in adolescent mental health trajectories across the digital era, although interpretations remain contested. A new generation of relational artificial intelligence (AI) systems—conversational and companion-oriented systems—may represent the next transition. Relational AI systems encompass both conversational platforms (eg, ChatGPT), primarily designed to assist with information seeking, learning, and everyday dialog, and companion-oriented platforms (eg, Replika), which are explicitly structured to simulate friendship, emotional support, or romantic connection through sustained interaction. These systems rely on persistent dialog, adaptive language generation, and varying degrees of conversational memory to maintain continuity across exchanges. Unlike earlier chatbots that responded to isolated prompts, contemporary relational AI systems often adopt personas, mirror emotional tone, and encourage ongoing engagement. Adolescents are among the fastest adopters, frequently using conversational systems for emotional disclosure, identity exploration, and social connection. For many young people, these technologies function not merely as tools but also as relational environments. However, their rapid diffusion has outpaced longitudinal evidence regarding developmental safety, raising questions about whether widespread adolescent exposure should be understood as a population-level health risk rather than solely a technological innovation. Emerging reports describing suicides associated with chatbot engagement, emotional dependency, and AI-reinforced manic or delusional ideation have drawn increasing attention from clinicians and researchers. These cases do not establish causal relationships between AI use and mental health outcomes. However, public health responses rarely wait for definitive proof when exposure is widespread and the population affected is developmentally vulnerable. Adolescence represents a period of heightened social sensitivity and ongoing neurodevelopment, during which digital environments can shape identity formation, emotional regulation, and perceptions of support. Technologies that simulate social relationships therefore warrant particular scrutiny. This Perspectives and Controversies article argues that adolescent exposure to relational AI systems should be understood through a public health lens. We propose that relational AI systems constitute a new category of developmental exposure, characterized by sustained simulated relationships rather than discrete technological interactions. Drawing on developmental science, sociotechnical risk analysis, and regulatory precedents such as Australia’s social media age restrictions, we suggest that relational AI may warrant age-appropriate design limits or limited deployment during adolescence. If adolescent mental health is a defining challenge of the digital era, then relational AI systems must be evaluated not only as innovation but also as a potential population-level exposure requiring the same caution applied to other emerging public health risks.