Abstract
Artificial Intelligence (AI)—especially Large Language Models (LLMs)—has become increasingly important in higher education, offering applications ranging from personalized learning paths to administrative process automation. Despite these advancements, there is a clear need for integrative frameworks that address both technical efficiency and ethical standards, particularly in data-sensitive academic contexts. This paper presents findings from a systematic literature review of 111 relevant studies (2014–2024) identified in Web of Science, Scopus, and Google Scholar. Guided by PRISMA methodology, we synthesized the potential benefits (e.g., adaptive feedback, administrative automation) and significant challenges (e.g., privacy issues, algorithmic bias, lack of model explainability). We propose three essential pillars for responsible and effective AI integration in higher education:
- Institution-owned AI systems to ensure robust data governance,
- Continually updated policy guidelines emphasizing transparency and academic integrity, and
- High-quality professional development for faculty to build both technical and ethical capacities.
Our findings underscore the need for holistic approaches that unite technical innovation with ethical accountability, thus facilitating sustainable, equitable, and data-protective AI adoption in universities. The conclusions and recommendations presented here provide a practical roadmap for institutions seeking to optimize learning processes while maintaining the highest standards of fairness and privacy.