The paper studies the Digital Horizon (DiHor) project, funded by the European Commission under the Erasmus+ programme, which is implemented by a partnership of universities, high schools, and NGOs specialising in education and IT from Italy, Portugal, France, and Romania. The DiHor project equips teachers, students, and parents with the knowledge, skills, and strategies necessary to successfully incorporate digital technologies and artificial intelligence (AI) into education and ensure students’ holistic human development.

The paper describes the bottom-up strategy underpining the project online course which will be tested with teachers, parents, and students. We asked all 78 students from EuroED High School Iasi, Romania about their AI usage and how this affects them. In designing the survey, we started from Matusov‘s (2016) mapping concepts of agency in educational contexts and proposed a framework covering the main domains of student agency in an AI-assisted environments: epistemic, strategic-operational, motivational or agentic, relational-collaborative, and ethical-critical.

We found that students use AI for study support, explanations, information, and revision, particularly in science. Still, many demonstrate active agency by seeking online confirmation, revising AI results, and looking for bias or missing views. The findings also show significant tensions, with over 50% not disclosing AI use in assignments, and few considering fairness, bias, or authorship. AI prejudice is underestimated, while collaborative regulation is weak since many students don't discuss AI in group work. Some students show a developing awareness of boundaries and appropriate use by avoiding AI when testing knowledge, or to protect creativity and maintain fairness.

The findings informed the school's AI strategy by aligning it with last year’s initiative when we created a dedicated AI-strategy committee and had two staff and the director attend the EDEN DLA event in Barcelona to discuss AI implementation in schools.