Conference Agenda

Governance is a no-man's land for adolescents ages 13-17. They occupy a gray area, too old for Children's Online Privacy Protection Act (COPPA) protections, too young for adult treatment. Despite this, they must navigate an information landscape that treats them as adults. COPPA is a U.S. law that protects the privacy of children under 13 online. Reintroduced in March 2025 COPPA 2.0 extends these protections through age 17. However, this does not address the urgency for governance of AI-mediated content delivery. Adolescents can use generative AI tools such as ChatGPT, Gemini, and the AI features of social media platforms like Snapchat to obtain personal health information in environments not designed for them. They often do not understand the output or the implications of the data they input. This analysis examines the ethical obligations of AI platform designers, health information providers, and federal regulators toward adolescents who use AI-mediated systems to seek, evaluate, and act on health information. Anchored by three pressing questions: (1) Do ethical harms and guardrails arise with AI-mediated adolescent health information content? (2) How do governance frameworks fail adolescents, given the 2025 COPPA Rule amendment, COPPA 2.0 (Markey, E. et al., 2025), and AI disclosure law? (3) How does empirical evidence about AI trust, information avoidance (Golman, 2017), and health equity inform governance? The gap above COPPA and below protection draws on a currently in-progress convergent mixed-methods study with adolescents ages 12-18 across two educational sites, composed of 2 charter schools and 1 community program serving under-resourced adolescents in the U.S. Data collect includes three instruments, an online survey (n=160-200, ages 12-18) using validated eHealth, ICT self-efficacy, with information seeking behavior and two constructs were developed to investigate AI Label Trust, the degree to which Al mediated content attribute to shaping adolescent credibility in judgement and Social Trust Bias, the degree to which likes signal peer engagement, share, and follow counts that override evaluation literacy when evaluating content. Semi-structured focus groups (n=15-20, ages 12-18) and adult key informant interviews (n=10-15, educators) provide context for the lived experience behind trust and avoidance decisions. Expanding on how adolescents rationalize unlabeled AI content and navigate charged health topics. This convergent design integrates quantitative and qualitative instruments, with policy implications, and is synthesized across both. This paper operates at the intersection of communication law and policy, AI ethics and governance, information science, and health equity research. It draws on principles of rulemaking in administrative law and children's rights scholarship in communication policy. The analysis uses Wilson's (1981) information-behavioral model, Sieck's (2021) Digital Health Equity Framework, and Golman's (2017) information-behavioral theory. A cross-disciplinary approach is taken to address the documented governance failures. This research investigates the behavioral and epistemic consequences of ungoverned AI-mediated health information environments and proposes equity-centered federal interventions. These ungoverned digital environments were not built with adolescents in mind for use. Those who most need reliable health information are often the ones who distrust it, avoid it, or accept social- and algorithmically amplified misinformation. This happens not simply due to a lack of intelligence or digital literacy, but because the system lacks a framework for young people to navigate this new landscape in a protected way. Three key findings are anticipated with direct governance implications. First, information avoidance as a mediator of the relationship between anxiety and information seeking, viewing disengagement as a policy-produced outcome rather than a literacy inadequacy (Golman, 2017). Second, AI Label Trust operates in two ways: labeled AI content does not automatically confer trust, which challenges disclosure of the solution assumption under current California chatbot law (2025). Third, Social Trust Bias is an equity-stratified predictor of credibility assessment, with adolescents in under-resourced settings who are algorithmically susceptible independent of individual literacy (Stevens, 2017). Collectively these findings builds four governance recommendations: (1) classify AI health systems as high risk, mandating pre deployment equity impact assessments for systems accessibility; (2) design AI disclosure standards separate form adults; (3) prohibit engagement optimization that ban platforms form using click bait algorithms for health content served to users under 18; and (4) create secure age appropriate environments that include private rights of actions for minors age 13-17 harmed by AI health content and mandate participatory governancethat includes adolescents as co designers of governance frameworks. In 2023, the U.S. Surgeon General identified an adolescent social media and mental health crisis; this issue extends to the evolving AI ecosystem in health information. Continued federal inaction on adolescent-specific protections leaves those aged above COPPA eligibility and below legal adulthood exposed to significant risks.