<p>The rapid proliferation of artificial intelligence-based large language models (LLMs) in everyday digital environments is fundamentally transforming how adolescents communicate, learn and express their emotions. While these systems offer accessibility and personalised interaction opportunities, their anthropomorphic designs, constant availability, and capacity to generate responses that mimic empathy pose significant ethical risks for adolescents, who are a developmentally vulnerable group. As adolescents have not yet fully developed their cognitive, emotional, and psychosocial autonomy, they may become more susceptible to emotional dependency, manipulation, and harm in interactions with human-like artificial intelligence. This study examines adolescents’ interactions with generative artificial intelligence language models within a bioethical framework, based on the publicised case of Adam Raine, a 16-year-old who lost his life following prolonged interaction with an artificial intelligence chat system. Through a principle-based ethical analysis grounded in the principles of respect for autonomy, beneficence, non-maleficence, and justice; it discusses how design features such as anthropomorphic language use, constantly affirming responses, personalisation, and the absence of effective crisis intervention mechanisms erode adolescent autonomy, produce foreseeable harm, and create disproportionate risks for vulnerable groups. The study argues that the Adam Raine case should be viewed not as an exceptional tragedy, but as an early warning sign of structural risks in current AI design and governance approaches, and that these risks can be mitigated through binding design measures, mandatory safety mechanisms, and governance/accountability frameworks.</p>