Generative Artificial Intelligence Policy: A Qualitative UNESCO Framework Analysis

Synthesis: Adarkwah et al. (2026) conduct a qualitative policy analysis of Generative AI (GenAI) policies at 30 highly ranked universities (by 2024 QS rankings) across the top ten countries for AI preparedness, evaluating them through UNESCO's eight-component GenAI framework. The findings reveal significant disparities in policy robustness: core ethical and AI Governance principles are widely embraced, yet critical issues of inclusion, equity, and Sustainability — such as internet access, gender parity in AI, and environmental impact — are often overlooked. Nordic countries and New Zealand cover UNESCO's elements more fully than some higher-ranked AI Preparedness Index (AIPI) countries, demonstrating that AIPI ranking does not guarantee strong GenAI Educational AI Policy. Notably, no public policies were found for German universities or Tallinn University of Technology. The study urges higher education leaders to develop more inclusive, future-oriented Higher Education policies integrating social equity, interdisciplinary experimentation, and sustainability.

Key Findings

• Uneven policy robustness: There are significant disparities in how leading universities regulate GenAI; while core ethical and AI Governance principles are broadly adopted, many policies remain declarative rather than operationally assured.
• Neglect of inclusion and sustainability: Key UNESCO elements such as inclusion, equity, internet access, gender parity in AI, and environmental impact are frequently overlooked across policies.
• AIPI ranking is not predictive: Nordic countries and New Zealand cover UNESCO's elements more fully than some higher-ranked AI Preparedness Index (AIPI) countries, showing that national AI readiness does not guarantee robust institutional policy.
• Missing policies: No publicly accessible GenAI policies were found for German universities or Tallinn University of Technology; four of the 30 targeted universities were excluded for lacking public policies, leaving 26 in the final analysis.
• Threat-focused framing: Reflecting broader Academic Integrity literature, many policies anchor in misconduct prevention and originality concerns, narrowing the pedagogical debate and under-addressing socio-political and moral dimensions.
• Recommendation: Higher education leaders should develop more inclusive, future-oriented policies integrating social equity, interdisciplinary experimentation, and sustainability into Educational AI Policy.

Study Design & Method

The study employed a qualitative policy analysis using the European Training Foundation's four-step guide (framing the problem, collecting/describing evidence, interpreting evidence, formulating recommendations). A purposive sampling strategy selected the top ten countries for AI preparedness based on the IMF's AI Preparedness Index (AIPI), then the top three world-class universities per country using the 2024 QS global rankings.

A total of 159 documents (policy texts, institutional announcements, webpages, reports) were retrieved and analyzed via content analysis with open coding. Two researchers independently coded policies for two countries (six universities) and met to establish consensus before broader team review. Institutional-level GenAI policies were evaluated against UNESCO's eight-component GenAI framework, which spans: inclusion and cultural-linguistic diversity; human agency; monitoring and validation; learner competencies; educator and researcher capacity; pluralism and epistemic diversity; local experimentation and evidence accumulation; and long-term, interdisciplinary, intersectoral review. An indicator was deemed neglected if addressed by five or fewer universities. Data collection ran October 2024 to January 2025.

What this means for practice

• Administrators. Audit your GenAI policy against all eight UNESCO components, not the ethical principles alone: inclusion and cultural-linguistic diversity, human agency, monitoring and validation, learner competencies, educator and researcher capacity, pluralism and epistemic diversity, local experimentation, and long-term interdisciplinary review — an indicator counted as neglected when five or fewer of the 26 analyzed universities addressed it.
• Administrators. Replace declarative statements with assurance structures: the analyzed policies were largely public-facing and anchored in misconduct prevention and originality concerns, so add the monitoring, validation, and review mechanisms that would make commitments enforceable.
• Administrators. Do not read national AI readiness or ranking as evidence of institutional policy strength: Nordic countries and New Zealand covered the UNESCO elements more fully than higher-ranked AI Preparedness Index countries such as the Netherlands and Germany, and no publicly accessible policy was found for the German universities or Tallinn University of Technology.
• Faculty developers. Build educator and learner GenAI competency programs as a named policy element. The UNESCO framework includes educator and researcher capacity, and the analysis found competency development weakly covered relative to ethics and academic integrity.
• Administrators. Write inclusion and sustainability into the policy text itself: internet access, gender parity in AI, reskilling for older and non-traditional learners, and environmental impact were among the weakest-covered elements, so moving beyond checklist compliance means naming them explicitly.

Limitations

• The purposive sample is the top three universities in the top ten countries on the IMF AI Preparedness Index (30 institutions targeted); four universities were excluded for lacking publicly accessible policies, leaving 26 in the analysis, and the authors state this ranking-based selection limits generalizability and may not identify the most progressive or well-informed institutions.
• Almost all analyzed universities are Western, with only Singapore outside that group, so the authors describe the findings as a Western and Eurocentric view of GenAI policy that does not provide a universal discourse.
• The evidence base is publicly available documents: reliance on public sources may overlook internal or evolving policies, and language barriers and limited institutional transparency further constrain access (the authors cite University of Copenhagen and TU Denmark policies kept on internal portals).
• The analysis is a documentary snapshot: data collection ran from October 2024 to January 2025, coding 159 retrieved documents and 100 eligible websites, with reliability established by distributed coding among four researchers and consensus of at least three rather than a reported reliability statistic.

Citation

Adarkwah, M. A., Ercan, A. M., Schneider, K., & Bayar, E. (2026). Generative Artificial Intelligence Policy: A Qualitative UNESCO Framework Analysis. Journal of University Teaching and Learning Practice, 23(2).