ABSTRACT Generative AI is transforming how youth interact with technology, introducing novel risks that are not fully understood by current safety frameworks. This poster presents an empirically grounded taxonomy of risks youth face in interactions with generative AI, based on analysis of 344 chat transcripts, 30,305 Reddit discussions, and 153 real‐world AI incident reports. We identify six main risk domains—mental wellbeing, behavioral and social development, toxicity, privacy, bias and discrimination, and misuse or exploitation—encompassing 84 specific risks across four interaction pathways. We highlight GAI‐unique risks such as emotional dependency, blurred boundaries between virtual and real life, and harmful GAI‐initiated behaviors. Our taxonomy offers a structured foundation for practitioners, educators, and policymakers to recognize and mitigate these emerging threats and bridge critical gaps in existing approaches to youth online safety and AI risk.
Yu et al. (Wed,) studied this question.