As Europe moves to formalise the ethical contours of artificial intelligence, the question of AI in education stands as both a promise and a provocation. At a recent high-level roundtable hosted by CroAI and Saviesa, an international panel gathered to examine this intersection—where regulation, innovation, and pedagogy meet.
Moderated by Peter Stojanovic, Editor at HotTopics and an experienced facilitator of global leadership dialogue, the panel brought together voices from legal, educational, and institutional domains. Among them were Stefan Martinić, a Croatian lawyer with expertise in technology and data protection law, and Leonor Diaz Alcantara, founder of Saviesa and a recognised leader in education and social impact. They were joined by representatives from CARNET, Croatia’s national centre for education technology, as well as educators and policymakers engaged in shaping the future of learning.
At the core of the discussion was the EU AI Act—groundbreaking legislation that categorises educational AI as “high-risk” due to its potential to influence life-altering decisions in grading, assessment, admissions, and behavioural monitoring. While its intent is to protect, some participants cautioned that excessive regulatory complexity could disincentivise experimentation within education systems already navigating rapid change.
The Croatian experience offered a compelling model. With over 300 AI-driven features implemented across its school system, Croatia is not theorising about the future—it is actively testing it. Initiatives such as Brain demonstrate that when AI is introduced through iterative, research-informed methods—with active involvement from educators and strong institutional backing—it can complement, not compete with, human teaching.
The discussion, however, transcended policy. It struck at the heart of a cultural shift: the need to safeguard critical thinking in a world increasingly influenced by black-box systems. AI literacy must evolve beyond operational skills to include judgment, ethical discernment, and the capacity to interrogate algorithmic outputs. Without this, technology risks displacing not just jobs—but reasoning itself.
Participants emphasised that education must not only teach about AI, but around it—embedding creativity, empathy, and collaboration into curricula alongside digital competence. These are not decorative additions. They are the human architecture required to navigate an automated world.
While the EU AI Act introduces important guardrails—including regulatory sandboxes that allow for controlled, real-world experimentation—its success will depend entirely on how it is implemented by member states. Frameworks alone are insufficient. It is in classrooms, ministries, training centres, and community conversations that the true impact will be determined. The roundtable closed with a shared understanding: AI is not a distant threat or future. It is a present tool. The challenge is no longer whether to use it, but how—with wisdom, with equity, and with care.
