Faculty occupy a peculiar position in higher education’s current rush to govern artificial intelligence. They are, simultaneously, the group most directly affected by AI policy, the group most responsible for translating institutional expectations into classroom practice, and the group least likely to have been meaningfully involved in writing the policies they are expected to follow.
As a result, faculty are being asked to implement AI policy they did not help write, enforce expectations they were not given, and navigate risks — to academic integrity, data privacy, and intellectual property — largely on their own. They are expected to set course-level AI guidelines without institutional consensus on what those guidelines should include. They are held accountable for student misuse of tools that their institutions have neither approved nor restricted. They are required to make judgment calls about data privacy in real time, often without knowing which AI platforms their institution considers safe, which violate FERPA, or whether anyone has asked those questions at all. And they are doing all of this while absorbing the message, implicit or explicit, that AI adoption is both urgent and inevitable — without being given the time, training, or voice in institutional decision-making that would make responsible adoption actually possible.
In short, there is a delegation of institutional responsibility onto the people least positioned to absorb it.
The Policy Gap Is Larger Than It Appears
In 2024, Inside Higher Ed surveyed chief academic officers across the country and found that only 20% of colleges and universities had published a policy or policies governing the use of AI, including in teaching and research (as cited in AAUP, 2025). In AI terms, two years is a long time. The landscape has shifted — though perhaps not in the ways institutions might hope.
By September 2025, a UNESCO survey of 400 higher education institutions across 90 countries found that 19% had a formal AI policy in place, with a further 42% reporting that frameworks were under development — meaning roughly 61% were at least actively engaged in the process (UNESCO, 2025). Progress, but uneven. Institutions in Europe and North America are considerably further along than those in other regions, largely due to regulatory and funding differences. And the awareness gap documented in 2024 has not meaningfully closed: 80% of faculty and staff report using AI tools, but fewer than 25% are aware of institutional policies (UNESCO, 2025) — nearly identical to the figures from the year prior. More institutions are writing policies. Fewer faculty know those policies exist.
What makes this more striking still is what sits alongside it: 90% of institutions in the same period reported introducing initiatives around AI for teaching, research, or work (AAUP, 2025). Initiatives without policies. Activity without accountability structures. The gap between institutional motion and institutional guidance is, for most faculty, the environment they are navigating every day.
The consequences are predictable. Azevedo et al. (2025) found that many faculty are required to navigate AI expectations independently — without institutional support, professional development, or cross-departmental alignment — and that policies, where they exist, vary widely in scope, enforceability, and the degree to which faculty were involved in their development. The result, as the research documents, is confusion about what is permissible, what is encouraged, and what is restricted, with real risks of unintentional policy violations by faculty and students alike.
The Policies That Matter Most to Faculty
When higher education institutions do develop AI policies with faculty in mind, four domains tend to carry the most operational weight.
Academic Integrity
Academic integrity is the domain where most institutions have moved fastest — and where the policy challenges are most visible. The shift from plagiarism detection to AI-generated content has fundamentally changed the nature of the problem. Traditional detection tools were designed to identify copied text. AI-generated content is, by definition, original, which renders those tools insufficient as standalone responses (Wu et al., 2024).
Institutions are responding in different ways. Oxford’s revised Academic Integrity Framework pivots away from detection-first approaches toward assessment redesign and transparent disclosure requirements (The Education Magazine, 2026). Columbia University’s guidance recommends that faculty set clear course-level expectations at the beginning of each semester, through the syllabus and direct classroom conversation, and treats any unauthorized AI use as equivalent to unauthorized assistance or plagiarism (Columbia University, n.d.). Many other institutions, including Johns Hopkins and Cornell, leave the determination largely to individual instructors — a flexibility that, in practice, creates the very inconsistency that leaves both faculty and students uncertain about what the rules actually are (Thesify, 2025).
The deeper issue is that academic integrity policies focused narrowly on detection or prohibition do not answer the harder question: what does authentic learning look like in a world where AI is widely available and widely used? Faculty need policy frameworks that engage that question, not ones that defer it.
Data Privacy
Data privacy may be the least visible AI policy issue for faculty, but it carries some of the most significant institutional risk. When faculty or students enter course content, unpublished research, or student information into public AI platforms, they may inadvertently expose data that is protected under the Family Educational Rights and Privacy Act (FERPA) or other applicable laws — in some cases without any awareness that a privacy boundary has been crossed (Columbia University, n.d.; The Education Magazine, 2026).
The AAUP’s 2025 survey found that at least 95% of faculty respondents identified maintaining data privacy as an area of critical importance in the context of AI policy (AAUP, 2025). Yet clear institutional guidance on which AI tools are approved for use with student data, what data may or may not be entered into AI platforms, and how faculty should navigate these questions in practice remains rare. The absence of this guidance does not reduce the risk. It distributes it — onto individual faculty members who may not have the technical or legal expertise to assess it.
Intellectual Property
Intellectual property is an area where faculty concerns are particularly acute and institutional policy is particularly underdeveloped. The questions are real and still largely unresolved: Who owns content produced with AI assistance? What happens to faculty research when it is processed by an AI platform whose terms of service grant the company rights to input data? What are the implications for peer review, publication, and scholarly attribution?
AAUP (2025) found that protecting intellectual property rights was among the areas of highest concern for faculty respondents, at or above the 95th percentile of importance. Yet few institutions have developed clear policies addressing these questions in the context of AI specifically. Columbia University’s guidance notes that AI outputs may violate the intellectual property rights of others and may not themselves be protected under intellectual property law — an acknowledgment of the complexity that stops short of providing faculty with actionable guidance (Columbia University, n.d.).
Academic Freedom and Shared Governance
Perhaps the most foundational concern — and the one most at risk of being treated as peripheral — is the question of faculty voice in AI governance itself. The AAUP’s 2025 national survey of 500 members across nearly 200 campuses found that 71% of respondents said administrators overwhelmingly lead conversations about AI in teaching, research, policy, and professional development, while gathering little meaningful input from faculty, staff, or students (AAUP, 2025). Eighty-five percent of respondents said they were at least somewhat concerned about how educational technology, including AI, is being implemented at their institutions.
This matters beyond questions of process. Under the AAUP’s longstanding principles on shared governance, faculty bear primary responsibility for determining appropriate curriculum and procedures of student instruction — a responsibility that explicitly extends to AI and other educational technology infrastructure (AAUP, 2025). When AI tools are adopted, mandated, or embedded in institutional systems without meaningful faculty involvement, the result is not just a governance failure. It is a potential infringement on the conditions that make academic freedom meaningful in practice.
What Faculty Are Experiencing
The data on faculty experience is difficult to set aside. AAUP’s 2025 survey found that AI implementation has led to at least somewhat worse outcomes for the teaching environment (62% of respondents), job enthusiasm (76%), and academic freedom (40%) (AAUP, 2025). But those numbers do not fully convey what is underneath them — because AI governance is not arriving in a vacuum. It is landing on a workforce that was already stretched before generative AI became a policy problem.
A systematic review of 60 studies covering more than 43,000 university faculty identified excessive workload, lack of institutional support, and workplace conflict as the three dominant drivers of academic burnout — a condition that has become, by any reasonable measure, endemic in higher education (Cadena-Povea et al., 2025). The Anthology 2025 Faculty Survey, drawing on responses from more than 2,500 US-based instructors, found that six in ten reported their workload had increased over the prior five years, with course creation, assessment design, grading, and student mental health support each consuming a growing share of an already compressed working day (Anthology, 2025). This is the environment into which AI governance expectations are being introduced. The cumulative effect matters. Each new initiative compounds the last. Faculty who have absorbed a decade of institutional change — online learning transitions, pandemic pivots, enrollment pressures, new accreditation requirements — are not encountering AI policy as a fresh challenge on a blank slate. They are encountering it as one more demand in a long sequence of demands, most of which arrived with urgency and left without resolution.
How faculty typically respond to this kind of accumulation is well documented, and it is not primarily resistance. It is triage. Faculty prioritize what feels immediate, what is measured, and what connects to something they already care about. Institutional mandates that do not meet any of those criteria tend to generate the surface compliance that mandates can produce — a syllabus statement copied from a template, a box checked on a professional development form — without generating the understanding or commitment that meaningful behavioral change actually requires. Required AI training is a particular case in point. When professional development is mandated without meaningful connection to faculty’s own teaching questions, without disciplinary relevance, and without adequate time allocated in the workday to actually engage with it, the training becomes another source of resentment rather than a source of support (AAUP, 2025). Attendance is not adoption. Completion is not competence.
The situation is most acute for adjunct and contingent faculty, who now constitute the majority of the instructional workforce in US higher education. Adjuncts typically receive no compensation for professional development, hold appointments that are renewed semester by semester, teach across multiple institutions simultaneously, and have no meaningful access to shared governance processes. They face the same course-level AI policy decisions as their full-time colleagues — the same ambiguities about academic integrity, the same data privacy risks, the same student questions — but without the institutional standing, job security, or support structures that might make those decisions navigable. The AAUP’s 2025 survey found near unanimity among part-time faculty that AI implementation has made their working conditions worse (AAUP, 2025). That finding is not surprising. It is the predictable result of asking the most structurally precarious members of the academic workforce to absorb institutional change with no resources and no runway.
These are not the responses of faculty who are resistant to change for its own sake. They are the responses of professionals who are being asked to implement something consequential with inadequate guidance, inadequate support, and inadequate voice in the decisions that affect them. As one AAUP respondent put it: institutions are making technology decisions that massively impact the teaching and research mission — and apparently no one thought to consult the faculty (AAUP, 2025).
From Policy on Paper to Policy That Holds
Azevedo et al. (2025) draw an important distinction between policy that is faculty-directed and policy that is faculty-inclusive. The first tells faculty what to do. The second builds the policy with them. That distinction matters, but it is also only the starting point. Even genuinely collaborative policy development can produce documents that are too broad, too static, or too disconnected from disciplinary reality to function in practice.
Effective faculty-centered AI policy has to operate at multiple levels simultaneously. At the institutional level, it must establish clear principles — on approved tools, data boundaries, academic integrity standards, and intellectual property — that give faculty a defined framework within which to exercise professional judgment. At the disciplinary level, it must allow flexibility to reflect the genuine differences in how AI intersects with a chemistry course versus a creative writing seminar versus a nursing clinical. A blanket prohibition on AI-generated content may make sense in one context and actively undermine learning in another. Policies that do not account for this place faculty in a position that Jiang et al. (2025) describe as a persistent implementation failure: when institutional guidance cannot be mapped onto the realities of a specific discipline or classroom context, faculty either ignore it or interpret it inconsistently — which is precisely the fragmentation that governance is supposed to prevent.
The sequencing of policy development matters in ways that institutions frequently underestimate. When people do not understand why a change is happening before they are told how to comply with it, compliance remains shallow. Faculty who receive AI policy as a finished document — without having been part of the conversation that shaped it, and without a clear articulation of the problem it is designed to solve — are unlikely to internalize it in any meaningful way. The policy gets filed, referenced when a specific student situation forces the question, and otherwise set aside. Institutions that have built more durable AI governance frameworks tend to have invested early in explaining the stakes — for students, for institutional liability, for the integrity of the academic enterprise — before asking faculty to adopt specific practices (Wu et al., 2024). That sequencing is not procedural courtesy. It is the difference between a policy that faculty own and one they merely tolerate.
Policy also needs to be treated as a living structure, not a deliverable. The HEAT-AI framework, developed through expert interviews and faculty stakeholder engagement at a European university, was built explicitly around the recognition that no governance document written today can anticipate the tools or use cases that will emerge in the next eighteen months — and that adaptive, regularly evaluated policy is the only kind that can remain functional over time (Temper et al., 2025). This has direct implications for how institutions resource governance. A policy that is published and then left static is not governing anything. It is a record of institutional intention at a moment in time. Real governance requires ongoing investment: regular review cycles, structured channels for faculty feedback, and a clear process for updating expectations as the technology and the evidence base evolve.
Professional development is the mechanism through which policy is most often operationalized for faculty — and it is where the gap between what institutions offer and what faculty actually need is most visible. The problem is rarely the absence of professional development opportunities. It is the nature of what is offered. Generic AI literacy training, delivered in a single session to a mixed audience of faculty from across disciplines, does not equip a political science professor to evaluate whether an AI tool is appropriate for a specific research assignment, or help a nursing faculty member explain to students why clinical documentation cannot be AI-assisted. What faculty need is professional development that is discipline-specific, anchored in real teaching scenarios, and paced in a way that does not add to an already unsustainable workload. Azevedo et al. (2025) are explicit on this point: critical AI literacy — the capacity to evaluate AI’s social, ethical, and disciplinary implications, not merely its technical functions — must be a sustained institutional investment, not a one-time training event.
Underlying all of this is a structural question that faculty-centered policy cannot sidestep: who is accountable for making sure the policy actually changes behavior, and what reinforcement systems are in place to support that change? Institutions that publish governance frameworks but leave implementation entirely to individual faculty — without visible leadership modeling, without consequences for non-compliance, and without recognition for faculty who are doing this work well — should not be surprised when adoption remains uneven. Policy without reinforcement is aspiration. The institutions building AI governance that lasts are the ones treating it as an organizational commitment, not a document.
References
- American Association of University Professors. (2025). Artificial intelligence and academic professions. https://www.aaup.org/reports-publications/aaup-policies-reports/topical-reports/artificial-intelligence-and-academic
- Anthology. (2025). Addressing the hidden crisis: The realities of faculty burnout. https://backstage.anthology.com/sites/default/files/2025-08/AddressingTheHiddenCrisisTheRealitiesOfFacultyBurnout_WhitePaper_v1.pdf
- 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
- Cadena-Povea, H., Hernández-Martínez, M., Bastidas-Amador, G., & Torres-Andrade, H. (2025). What pushes university professors to burnout? A systematic review of sociodemographic and psychosocial determinants. International Journal of Environmental Research and Public Health, 22(8), 1214. https://doi.org/10.3390/ijerph22081214
- Columbia University Office of the Senior Vice Provost. (n.d.). Generative AI policy. https://provost.columbia.edu/content/office-senior-vice-provost/ai-policy
- EDUCAUSE. (2024). 2024 EDUCAUSE action plan: AI policies and guidelines. https://library.educause.edu/resources/2024/5/2024-educause-action-plan-ai-policies-and-guidelines
- Jin, Y., Yan, L., Echeverria, V., Gašević, D., & Martinez-Maldonado, R. (2024). Generative AI in higher education: A global perspective of institutional adoption policies and guidelines. Computers and Education: Artificial Intelligence, 8, Article 100348. https://doi.org/10.1016/j.caeai.2024.100348
- The Education Magazine. (2026, March 6). AI governance in higher education: The 2026 framework for policy & risk. https://www.theeducationmagazine.com/ai-governance-in-higher-education/
- UNESCO. (2025, September). UNESCO survey: Two-thirds of higher education institutions have or are developing guidance on AI use. https://www.unesco.org/en/articles/unesco-survey-two-thirds-higher-education-institutions-have-or-are-developing-guidance-ai-use
- 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
