Competent but Concerned: How AI Competence Shapes Students' Perceptions of AI-Related Risks in Upper Secondary Education

Students are the primary users of AI tools in classrooms, yet their perceptions of AI's risks remain largely absent from the research that shapes how these tools are introduced into education. This research examined the relationships between Finnish upper secondary students' (N=1525) self-perceived AI competence and their AI-related risk perceptions. Students reported their self-perceived AI competence and concerns about AI across the systemic, institutional, and personal domains. At the item level, higher competence was associated with reduced endorsement of nearly all concerns, with one notable exception: concern about AI-facilitated cheating increased with competence. The co-occurrence network analysis clarified this asymmetry: higher-competence students' concerns converged around a single, tightly organized cluster centered on the cheating concern, drawing together bias, inaccuracy, and reduced critical thinking, creativity, and learning, whereas lower-competence students reported broader and more numerous concerns, organized around resource use, copyright, and privacy issues that spanned systemic, institutional, and personal domains alike. The findings suggest that students' self-reported AI competence relates to both the number of concerns they express and the intertwined structure of their concerns about AI in education (AIED). The results highlight the need to incorporate AI literacy into curricula, provide teacher guidance, and inform policy development to ensure equitable integration of AI into K–12 education.