Yaman Yu, Yiren Liu, Jacky Zhang, Yun Huang, and Yang Wang, University of Illinois Urbana-Champaign
Generative AI is changing how youth engage with technology, yet the unique risks they face remain underexplored and are missing from existing safety frameworks. Without a focused taxonomy, important harms to youth may be overlooked. To fill this gap, we present the first Youth-Centered Risk Taxonomy for Generative AI, built from 344 youth–GAI chat logs, 30,305 Reddit discussions, and 153 AI incident reports. We identify six key risk categories and 84 specific risks organized along four interaction pathways. Our findings reveal new risks, e.g., Mental Wellbeing Risks, Behavioral and Social Developmental Risks, and new manifestations of Toxicity, Privacy, Bias/Discrimination and Misuse/Exploitation, which are not addressed in existing child online safety taxonomies and AI risk taxonomies. Grounded in real-world data, this taxonomy offers a clear framework to help AI practitioners, educators, parents, and policymakers better understand and address risks in youth–GAI interactions.
Open Access Media
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.
author = {Yaman Yu and Yiren Liu and Jacky Zhang and Yun Huang and Yang Wang},
title = {{Youth-Centered} {GAI} Risks ({{{{{{{YAIR)}}}}}}}: A Taxonomy of Generative {AI} Risks from Empirical Data},
booktitle = {Twenty-First Symposium on Usable Privacy and Security (SOUPS 2025)},
year = {2025},
isbn = {978-1-939133-51-9},
address = {Seattle, WA},
pages = {149--165},
url = {https://www.usenix.org/conference/soups2025/presentation/yu},
publisher = {USENIX Association},
month = aug
}
