RESPONSIBLE AI IN EDUCATION:

Abstract

The past three years have produced an accumulation of frameworks, guidelines, and regulatory instruments intended to govern the use of artificial intelligence in education. The EU AI Act, institutional AI policies, and national strategies have collectively established a vocabulary for responsible AI in educational settings. What they have not yet produced is a practical account of what responsible AI adoption actually requires from institutions navigating these frameworks under real operational conditions. This paper draws on experience in AI ethics research, strategic consulting, and the co-development of an AI-assisted academic assessment platform (MeritMinds) to examine the gap between responsible AI principles as written and responsible AI as practiced. The analysis focuses on three structural mismatches that current frameworks consistently underestimate: the gap between risk classification and institutional capacity to act on it; the gap between transparency requirements and the ability of faculty and administrators to meaningfully interpret AI outputs; and the gap between accountability language in policy documents and actual lines of responsibility when AI-assisted decisions cause harm. The paper proposes a practitioner-oriented diagnostic for educational institutions, structured around three levels of readiness — awareness, integration, and accountability — and offers specific questions institutions should be able to answer before deploying AI tools in high-stakes contexts. Particular attention is given to assessment and feedback tools, which the EU AI Act classifies as high-risk applications, and to the position of higher education institutions as simultaneously AI adopters, AI developers, and AI governance actors.