Spend any time in higher education right now and you will hear the word “governance” attached to artificial intelligence with increasing urgency. Committees are forming. Task forces are being convened. Policy drafts are circulating. And yet, for all the activity, there remains a striking lack of clarity about what AI governance actually is, what it is supposed to accomplish, and why — despite genuine institutional effort — getting it right has proven so elusive.
So What Is AI Governance?
At its most basic, AI governance refers to the policies, processes, structures, and norms that guide how artificial intelligence is developed, adopted, and used within an organization. In higher education, this typically encompasses decisions about which AI tools are permitted, how they may be used in teaching and research, what data protections apply, how academic integrity is defined and upheld, and who is responsible for oversight and enforcement (EDUCAUSE, 2024a; Wu et al., 2024).
That definition sounds manageable. In practice… well, it’s not.
At the international level, UNESCO’s 2021 Recommendation on the Ethics of Artificial Intelligence — the first global standard-setting instrument of its kind, adopted by 193 member states — establishes that AI governance must be grounded in human rights, transparency, fairness, and human oversight (UNESCO, 2021). These principles provide a useful ethical foundation. Yet translating global principles into operational institutional policy is a different undertaking entirely, and one that higher education institutions are only beginning to navigate.
The challenge is that AI governance is not a single policy problem — it is a cluster of overlapping and sometimes competing problems that touch nearly every function of an institution simultaneously. A decision about AI use in student-facing advising tools implicates data privacy, equity of access, and student affairs policy all at once. A faculty policy on AI in the classroom intersects with academic freedom, assessment design, and professional development. A procurement decision about an AI vendor involves legal compliance, IT security, and pedagogical implications that most procurement offices are not equipped to evaluate alone. In recognition of this complexity, EDUCAUSE (2024a) has proposed a governance framework organized around three interconnected domains: governance (including data governance, intellectual property, and equitable access), operations (including professional development and infrastructure), and pedagogy (including academic integrity, assessment practices, and student AI competency development).
In other words, AI governance requires institutions to govern something that does not fit neatly within any existing structure — and to do so under conditions of significant and rapidly evolving uncertainty.
Why Comprehensive Policy Is So Difficult
The Technology Moves Faster Than the Institution
Higher education institutions are not built for speed. Policy development is, by design, a deliberate process — one that involves shared governance, legal review, faculty input, and administrative approval. These are not inefficiencies. They are features of an institutional culture that values deliberation, consensus, and accountability.
Artificial intelligence, however, is not waiting. The tools available to students and faculty today are fundamentally different from those available eighteen months ago, and there is no reason to expect that pace to slow. A review of AI guidance at 14 major research universities in the United States found that many institutional guidelines, while well-intentioned, were developed rapidly and likely contained suggestions that had not been sufficiently tested in practice — and that many were already disjointed, hard to enforce, or quickly outdated by the time they were published (Wu et al., 2024). This isn’t a case against developing policy. It’s a case for approaching it differently – focusing on clear principles and adaptable processes instead of locking in rules for specific tools or use cases, and treating policy as something that evolves over time rather than something ever truly finished.
The Scope Is Genuinely Unprecedented
Most institutional policies govern something bounded — a process, a tool, a functional area. AI governance, by contrast, cuts across the entire enterprise. As Azevedo et al. (2025) observe, AI now touches admissions, financial aid, student support services, library resources, research compliance, and virtually every dimension of teaching and learning. An institution that tries to govern all of this through a single comprehensive policy is likely to produce a document that is either too broad to be actionable or too prescriptive to survive contact with reality.
The scope problem is compounded by the fact that different stakeholder groups — faculty, students, administrators, IT professionals, legal counsel, often have fundamentally different concerns, use cases, and levels of AI literacy. Faculty, in particular, are frequently left unsure about how to use AI tools in alignment with institutional values, especially when policies vary across departments and are inconsistently communicated (Chan, 2023, as cited in Azevedo et al., 2025). A policy written to satisfy all stakeholders simultaneously may end up serving none of them well.
Definitions Are Contested — and Consequential
What counts as AI? This sounds like a semantic question, but it is not. Whether an institution classifies a grammar-checking tool the same way it classifies a generative writing assistant has real implications for policy scope and enforcement. Whether “AI-generated content” is defined narrowly or broadly changes everything about how academic integrity policies apply. Notably, UNESCO’s (2021) own Recommendation on the Ethics of Artificial Intelligence deliberately avoids providing a single fixed definition of AI, recognizing that any such definition would need to evolve in accordance with technological development — a recognition that institutions developing static policies would do well to absorb.
These definition choices aren’t just technical details. They say something about an institution’s values—what it believes learning is for, what counts as academic work, and who gets to decide. When institutions skip over this groundwork to move quickly on policy, they often end up enforcing rules that don’t quite reflect what they intended.
Equity Considerations Are Frequently Underweighted
Perhaps the most consequential difficulty in AI governance is the tendency to treat equity as an afterthought — something to be addressed after the main policy architecture is in place. This is a mistake, and evidence from related domains suggests it is a costly one.
When institutions govern new technologies without an explicit equity lens, they tend to reproduce and amplify existing disparities rather than address them. As Azevedo et al. (2025) note, institutions risk exacerbating digital divides if AI integration is unevenly implemented, and students who independently discover how to use AI effectively gain significant advantages over those who do not — advantages that are not randomly distributed, but tend to follow existing lines of access and privilege. A governance framework that does not account for these differences is not neutral. It is inequitable by default.
Compliance and Buy-In Are Not the Same Thing
Even well-written policies fail when the conditions for implementation are not in place. Data from EDUCAUSE’s (2024b) research on AI in the higher education workplace reveals a striking gap: while 94% of higher education professionals reported using AI tools for work, only 54% were aware of any institutional policies or guidelines meant to guide that use.
This is not a compliance failure. It is an installation problem. Publishing a policy, standing up a task force, announcing a framework — these are acts of installation. They are necessary. They are not sufficient. And in higher education’s current push to govern AI, they are frequently where the work stops.
It is an awareness and buy-in failure — and it points to something that research on organizational change has documented consistently: the tendency to treat policy development as the finish line, rather than the starting line, is one of the most reliable predictors of implementation failure.
Mehl and Tsibolane (2025) further underscore this point, finding in their review of AI governance studies that restrictive policies tend to limit digital literacy while permissive policies foster over-reliance — and that guided frameworks, which require active stakeholder engagement, produce the most constructive outcomes. The implication is clear: governance that is done to faculty and students, rather than with them, is unlikely to achieve its intended effects.
What This Means Going Forward
AI governance in higher education is difficult not because institutions lack commitment or intelligence, but because the problem itself is genuinely hard. It requires governing something fast-moving and pervasive with tools built for slower, more bounded challenges. It requires building consensus across stakeholder groups with different interests and different levels of readiness. And it requires doing all of this while maintaining the institutional values — academic freedom, equity, shared governance — that make higher education distinctive.
None of this means comprehensive governance is impossible. It means that the path to effective governance runs through clarity about what governance is actually trying to accomplish, honest assessment of institutional readiness, and a sustained commitment to implementation — not just policy production. As EDUCAUSE (2024a) has argued, the goal is to move from reaction to action: from ad hoc responses to deliberate, coordinated governance structures that can evolve as the technology and our understanding of it continue to change.
The institutions that get this right will not be the ones that move fastest. They will be the ones that build governance structures grounded in their own values, attentive to the needs of their most vulnerable stakeholders, and designed to grow alongside the technology they are trying to govern.
Need Help?
References
- Azevedo, J., & colleagues. (2025). Institutional policies on artificial intelligence in higher education: Frameworks and best practices for faculty. New Directions for Adult and Continuing Education. https://doi.org/10.1002/ace.70013
- EDUCAUSE. (2024a). 2024 EDUCAUSE action plan: AI policies and guidelines. EDUCAUSE. https://library.educause.edu/resources/2024/5/2024-educause-action-plan-ai-policies-and-guidelines
- EDUCAUSE. (2024b). The impact of AI on work in higher education. EDUCAUSE. https://www.educause.edu/research/2026/the-impact-of-ai-on-work-in-higher-education
- Mehl, E., & Tsibolane, P. (2025). AI governance in higher education: Linking policy, cognitive effort, and academic confidence. ACIS 2025 Proceedings. https://aisel.aisnet.org/acis2025/83/
- Organisation for Economic Co-operation and Development. (2023). Emerging governance of generative AI in education. In OECD Digital Education Outlook 2023. OECD Publishing. https://www.oecd.org/en/publications/oecd-digital-education-outlook-2023_c74f03de-en
- UNESCO. (2021). Recommendation on the ethics of artificial intelligence. United Nations Educational, Scientific and Cultural Organization. https://unesdoc.unesco.org/ark:/48223/pf0000380455
- Wu, C., Zhang, H., & Carroll, J. M. (2024). AI governance in higher education: Case studies of guidance at Big Ten universities. Future Internet, 16(10), Article 354. https://doi.org/10.3390/fi16100354
