Key points:
- Governing AI requires an explicit architecture of differentiated authority
- When AI’s architects ask to slow down, education must speak up
- AI as a scaffold for learning: From access to judgment
- For more news on AI in higher ed, visit eCN’s AI in Education hub
If governance must preserve institutional purpose and leadership must remain accountable for institutional direction, an important question must be answered: Who possesses the authority to make the decisions through which that purpose is maintained and direction is determined?
In an AI-enabled world, a university could conceivably fail to prepare students even while everyone is focused on ensuring their preparation. Leadership may continue consultation without establishing clear expectations. Governing boards may seek to minimize risk while unknowingly increasing it through immobility. State and federal government agencies may add protections or restrictions. Information technology offices may prevent experimentation because no approved environmental environment exists, and procurement actions might embed a vendor’s assumptions into institutional processes without anyone else knowing. Academic programs may postpone curricular reconsideration, leaving students to encounter markedly different expectations from faculty across the same institution, and even the same department. Each action may be defensible within the domain of the actor taking it, but together they can determine institutional direction without that direction ever having been considered or authorized as a whole.
This is the most urgent unresolved challenge of AI governance. The issue is not simply one of distributing responsibility for action but of locating authority over action, or the lack thereof, defining where that authority ends, and ensuring that those who can delay, constrain, or prevent institutional action remain accountable for consequences extending beyond their domain. Because those consequences are borne by students, workplaces, the public and society, institutional direction cannot be left to emerge from disconnected decisions, unresolved conflicts and accumulated defaults. Governing AI therefore requires an explicit architecture of differentiated authority.
Participation is not authority
Historically, higher education has responded to issues of importance by broadening participation with the goal of improving judgment, better understanding consequences, and strengthening legitimacy through shared governance. This, however, does not determine who possesses the authority to make a binding decision. In similar fashion an administrative unit may be responsible for implementing an AI system without having authority to determine the institutional purpose it should serve. Institutional leaders may remain accountable for ensuring that graduates are prepared for an AI-driven workplace without determining how that preparation is expressed in every program or course. The central question, however, is not simply who participates, but who may decide, who may prevent a decision, who may reconsider it, and ultimately who answers when its consequences extend beyond the level at which it was made.
The authority to act, and not to act
Authority is most visible when something is approved, adopted, authorized or established. It is less visible, but no less consequential, when approval is delayed, resources remain unavailable, jurisdiction is left unresolved, or no viable path for responsible action is provided. Such negative authority is often necessary to protect aspects such as privacy, security, academic integrity, institutional obligations, and individual rights. But under no circumstances can this be viewed as neutrality. An unresolved curricular decision can result in students graduating without the critical preparation they need to succeed in a fast-changing workplace. Continued consultation without an eventual decision does not preserve the status quo since technology continues to advance, professional expectations change, and workplace needs further evolve while the institution, and students, wait, resulting in a direction being determined by default rather than institutional choice.
A decision attained by delay or inaction is no less consequential than one reached through approval or active denial. Those able to constrain or prevent institutional action must therefore be accountable not only for “harms” they seek to avoid, but also for the capabilities, opportunities and preparation their decisions may, and do, foreclose. The issue is not whether every refusal should be overturned, but rather whether authority over inaction is visible, traceable, and accountable when its consequences extend beyond the domain in which it is exercised.
From public obligation to institutional direction
While authority should reside at a level through which judgment can be exercised, the mere fact of established jurisdiction, organizational position or specialized expertise is not sufficient to legitimize the exercise of authority since it must be appropriate to the scope of the decision, responsive to consequences wider than just the immediate domain, and supported by an understanding of the conditions in which the consequences of the decision will unfold.
This architecture begins beyond institutional walls. State and federal governments establish statutory protections through legal and public obligations within which universities may use AI addressing aspects such as civil rights, privacy, accessibility, consumer protection, due process and public accountability. States are also justifiably establishing broad expectations and responsibilities of public institutions for educational attainment and workforce preparation. These regulations often also result in the exercise of negative authority in that a restriction intended to avoid one form of harm may prevent institutions from developing the capability to address another. Such public authority must therefore be accountable not only for the protection it creates, but also for the institutional and public possibilities it unnecessarily forecloses.
In similar fashion, the authority of governing boards extends to mission, risk, institutional capacity and assurance under the wider responsibility of translating public and fiduciary obligations into institutional oversight. While boards should require evidence that the institution has established decision rights, safeguards and the ability to respond to consequential change, they should not select systems, design detailed curricula or administer implementation. Risk aversion can itself become a consequential decision. Campus leadership converts public obligations and board-level expectations into institutional direction through the establishment of priorities, definition of required capabilities and safeguards, and allocation of resources. They cannot determine every academic or operational use of AI, and neither can they delegate institutional direction, or responsibility for seeing its cumulative effects, to a committee, information technology unit or academic body since none of these are independently accountable for the institution as a whole. Consultation can thus inform direction, but it cannot be a substitute for it.
From institutional direction to operational and academic practice
Institutional direction has meaning only when it becomes operational and becomes part of the academic ethos. Strategy, and by extension direction, are meaningless if the culture does not exist to enable it. This takes on critical emphasis in the context of AI where large, and significant, sections of institutions remain vehemently opposed to it and position it as the ruin of education irrespective of use or context. Faculty independence and student choice present the most difficult tests of governance and societal relevance of institutions because both can become forms of unilateral veto when extended beyond their appropriate scope.
In a balanced approach operational units determine the conditions under which AI can be implemented while academic units determine how institutional expectations become curricula, learning and demonstrated capability. While technical and operational requirements are necessary, control of systems, data, contracts and implementation cannot become authority over institutional purpose, and neither should units be able to alter or prevent institutional direction without formally claiming authority over it. The level of knowledge and competencies represented by a degree cannot be left to an accumulation of disconnected choices. If the institution determines that graduates require specific capabilities to work critically, responsibly and effectively with intelligent systems, those expectations must become coherent program outcomes. Authority over how learning occurs cannot become an unreviewable veto over whether essential preparation occurs.
Authority must remain answerable to relevance
Institutional failure can occur even when authority is allocated appropriately if it remains insulated from the world and the effects of external changes on internal direction and functioning. This is especially consequential in higher education, where deliberative processes and curricular changes have historically moved slowly, often over years rather than weeks or months. In addition, there is the gap between when decisions are made and when the affected students graduate, because a university does not prepare students only for the conditions visible when a curriculum is approved but more importantly for conditions that will exist when they graduate and that will continue changing throughout their careers.
A university must remain relevant to students who depend on education for socioeconomic mobility, workplaces and professions that will receive its graduates, the state and public that support it, and a society that relies on higher education to build capability and respond to change. These are not peripheral interests to be consulted after decisions are made but are fundamental, mission-driven, obligations that institutional decision-making cannot legitimately ignore and against which the exercise of institutional authority must be judged. Institutions have important constituencies that define relevance but do not have direct authority over institutional practice. Employers and professional bodies can make visible how AI is changing work, redistributing tasks and altering expectations of graduates, and alumni can further communicate how competencies are tested across careers and sectors. Students themselves can illuminate differences between intended effects of the curriculum and actual preparation for the workplace. Communities can make visible the needs of the region and can identify the human, civic and economic capabilities their regions increasingly require. When evidence indicates that an internally valid decision is weakening student preparation, regional capability or institutional relevance, the architecture must identify who has authority to require revectoring. Without this the institution may preserve legitimacy of process while losing relevance of results.
The final challenge
It must be emphasized that the absence of an institutional decision does not imply that no decision has been made but rather indicates that institutional direction has been ceded to whoever controls the default, the delay, or the veto. This is the more critical risk posed by an undefined architecture of authority, not that authority is located at the wrong level, but rather that formal authority remains undefined while consequential power is exercised through restrictions, implementation controls, unresolved decisions and accumulated local choices.
Leadership must make that authority explicit, establishing who may act, who may prevent action, where the authority ends and when consequences extending beyond one domain require reconsideration. Above all, it must ensure that authority over inaction carries accountability for what may be lost. The objective is neither unrestricted adoption nor institutional uniformity. It is the capacity to make timely, informed and accountable choices about where AI should be used, where it should be constrained and what human capability must be preserved or strengthened. Institutions that build that capacity will not merely react to the integration of intelligent systems but be able to determine deliberately how those systems should, and should not, shape their practices, their obligations and the preparation of those they serve.
Higher education’s culminating governance question is therefore not simply how AI should be used but is in deciding who has the authority to shape the institutional response, who has the power to prevent it, and who will answer for the future those decisions make possible, or place beyond reach. This is ultimately the test of governance in an age of intelligent systems. An institution that cannot locate authority cannot exercise judgment coherently. An institution that cannot connect authority to accountability cannot govern the consequences of its choices. Ultimately, an institution that allows its future to be determined through accumulated defaults, or vetoes, cannot remain intentionally relevant to the future unfolding beyond it.
Vistasp M. Karbhari is a Professor in the Departments of Civil Engineering, and Mechanical & Aerospace Engineering at the University of Texas at Arlington, where he served as President from 2013-2020.
- The data retention challenge for online exam platforms - October 28, 2026
- College was the default answer, but the data says it’s time to ask better questions - October 26, 2026
- The most advantaged students attend the best-funded colleges. Does it have to be this way? - October 23, 2026