From Restriction to Empowerment: Responsible AI Governance in Higher Education

Abstract

Generative AI tools have entered university life faster than institutions have been able to govern them, prompting policy responses that range from outright prohibition to largely unguided adoption. This paper examines the shift from restrictive to empowerment-oriented AI governance in higher education through a convergent parallel mixed-methods study comprising a structured survey (n = 71) and twenty-five semi-structured interviews across eleven academic disciplines at a private university in Bangladesh. The study documents patterns of AI adoption, stakeholder attitudes toward institutional policy, and barriers to responsible use. Participants situated within more empowerment-oriented institutional environments, including those characterised by AI literacy support, clearer ethical guidance, and stronger faculty engagement, reported greater confidence in using AI responsibly and clearer understanding of acceptable practices. By contrast, participants operating under restriction-only policies more often described uncertainty, confusion, and rule evasion. Regression analysis further showed that AI familiarity, frequency of use, and policy awareness were significantly associated with stronger support for empowerment-oriented governance. These findings inform a five-pillar framework for responsible AI integration encompassing AI Literacy Integration, Stage-Based Access, Transparent Use Norms, Assessment Innovation, and Faculty Development. Informed by stakeholder evidence and refined in dialogue with existing literature, the framework offers a practical model for institutions navigating AI governance in resource-constrained contexts. The paper contributes empirical evidence from a developing-country setting that remains underrepresented in current scholarship and highlights how responsible AI governance can support more equitable and context-sensitive higher education.