The answer is yes, though not for the reasons commonly assumed. AI itself isn’t dangerous; the real problem is that schools are using this powerful technology with old systems, like scattered data, disconnected tools, and teaching methods built for a time when information was hard to find. If these basics don’t change, AI won’t fix education. Instead, it will make current problems worse. The most urgent issue is not technology itself, but governance.
Adoption has already occurred, but governance has not
The numbers tell a consistent story throughout regions. According to the EDUCAUSE AI Landscape Study 2025, 57% of university leaders in the United States already identify AI as a strategic priority. That’s broad executive support. But the same study found a significant gap in execution: fewer than half of those institutions had a formal, comprehensive AI policy.
UNESCO’s global data confirms the pattern: only 19% of higher education institutions have an official AI policy, another 42% are still drafting one, and the rest have none.
The gap between fast AI adoption and slow governance is the main challenge for higher education in 2026. This issue isn’t limited to universities. Governments face it too, but on a bigger scale. Some avoid creating new regulatory bodies and prefer light, sector-specific rules. Others, like the European Union, have written risk categories into law. In Latin America, a public audit agency recently said that people, not AI systems, are always responsible for AI outputs. Different governments are making different choices, but the core problem is the same. Higher education can’t ignore it.
Our primary concern is not student use of AI, but the lack of institutional guidance, no clear rules, ethical frameworks, data protection, or shared pedagogical standards. Each teacher addresses these issues independently, and institutions are improvising. Improvisation at this scale leads to significant consequences, and an institution requires collective effort to change direction, as progress is limited if only a few contribute.
To move forward, the education strategy must evolve. It must be more than a document; it must reflect a commitment to changing how people work, learn, and lead together. That requires an anchoring strategy in the lived experience of schools and the building of conditions where people are not merely compliant but fully engaged in meaningful, coordinated change. This is not a minor detail; it is the foundation for all subsequent actions.
Three Challenges Institutions Continue to Face
These challenges are not new; AI has simply made them more urgent.
- Data is fragmented
AI only generates value when it’s fed high-quality data, and in most institutions, that data lives in silos: academic records in one system, payments in another, attendance in yet another, the student’s full history nowhere in particular.
However, building AI on top of a fragmented architecture doesn’t produce results; it magnifies the wrong things faster, as suggested in the World Bank report, Artificial Intelligence Revolution in Higher Education: What You Need to Know : An interconnected infrastructure enables institutions to deploy predictive models that detect at-risk students sooner and support proactive retention efforts. Without it, these insights and interventions remain out of reach.
Some institutions already show what it means to get organized first. Instead of relying on a single AI vendor or model, they build platforms that can run multiple models together, comply with data protection rules, and retain only aggregate usage data, not the details of each interaction. They didn’t start by buying an AI product. They first unified their data, then added AI on top.
The first step is not to purchase an AI tool, but to organize internal systems. We refer to this as a Connected Campus: a unified view of each student that enables further advancements.
- Governance doesn’t exist
Most institutions have no active stance on AI use, which doesn’t mean no one is using it; it means everyone is using it differently, with no shared criteria, no protection, and no transparency.
What happens when a teacher uploads student data into a free AI tool to grade exams? In most cases, that data trains an external model, not just an ethical risk, but in many jurisdictions, a violation of student rights. Some institutional directives say this plainly: tools like ChatGPT come with “no guarantees of personal and institutional data protection,” which is exactly why it makes sense to build an alternative of your own instead of leaving each teacher to freelance with public tools.
What happens when an admissions or grading system makes decisions without human oversight? The EU AI Act doesn’t leave that ambiguous: it classifies AI systems used to determine access or admission to education, evaluate learning outcomes, or detect prohibited behavior during exams as “high-risk,” which triggers mandatory audits, transparency obligations, and human-in-the-loop validation. Other institutions have arrived, independently and without any law requiring it, at the same red line: banning the use of generative AI for high-impact decisions, hiring, evaluating academic performance, resolving complaints, without first checking with the relevant legal and technology teams. Different institutions landing on the same line on their own is a signal, not a coincidence.
One practical framework gaining traction is the stoplight model: red means AI is prohibited, and any use is treated as plagiarism; yellow means it’s allowed with citation and submission of the prompts used; green means it’s required, the task was designed specifically to be completed with AI. Simple, visual, and easy to embed directly into the LMS next to each assignment.
- Pedagogy hasn’t been redesigned
This is the deepest problem, and the one that concerns us most. AI is now very good at three things that used to be exclusively human: generating ideas, structuring content, and expressing it clearly, the first three steps of classical rhetoric (inventio, dispositio, elocutio), as one university president put it to us recently. If AI does all three for the student, what’s left to teach?
The solution isn’t to ban AI. It’s to change what we measure. If exams only look at the final product, AI makes them useless with a single click. But if exams focus on the thinking process, the ability to question, check, and review what an algorithm creates, AI can’t replace that, because it can’t copy real human thinking.
The OECD’s Digital Education Outlook 2026 recommends universities teach students to “question with criteria”, auditing algorithm-generated outputs rather than trusting them by default. The World Bank adds a concrete practice: ask students to submit not just the final work, but the prompts they used and a reflection on their process. María Cristina Kanobel, researcher on education, takes this a step further and gives it a name, “augmented authorship”, defined as reflective co-creation between people and AI systems that requires meaningful human monitoring, disclosure of every algorithmic intervention, and demonstrable added intellectual value, not just a checkbox declaring “AI was used here.” (Notice how naturally this pairs with the stoplight model above.)
UNESCO’s Recommendation on the Ethics of Artificial Intelligence adopted by all 193 member states, the only AI ethics instrument with that level of consensus, says something worth quoting directly: AI systems used in learning “should be subject to strict requirements when it comes to the monitoring, assessment of abilities, or prediction of the learners’ behaviours,” and AI “should support the learning process without diminishing cognitive abilities.” A 2025 MIT study raised an uncomfortable possibility: when generative AI takes on too much of the cognitive work, learning can suffer. Participants who relied heavily on ChatGPT showed lower levels of neural engagement, memory retention, and critical thinking, suggesting that the tool may sometimes shortcut the learning process rather than reinforce it.
That’s the difference between using AI as a shortcut and using it as a co-pilot.
Teachers aren’t being replaced, they’re being repositioned
We hear a lot about teacher fear, and it’s not irrational. Technology moves fast enough that there’s barely time to learn what exists today before something new replaces it. Paralysis, in that context, is an understandable response. It’s also not unique to teaching; it shows up in every profession facing the same pace of change. But there are some things only humans can do: show empathy, make ethical decisions, connect with others, and understand a group’s mood. These are exactly what education needs more of, not less.
What’s changing is how teachers use their time. Passing on information isn’t the main job anymore, since information is everywhere and easy to access. Now, the focus is on helping students think, ask questions, and work together. These so-called “soft skills” are actually the hardest to teach and the hardest for AI to copy and automate.
Teachers who use AI tools save several hours a week on grading, planning, and admin tasks. But this only leads to better teaching if the whole system is redesigned. If a teacher saves time with AI but is still judged only on test scores, they can’t use that extra time for more important work. Real change needs the whole system to change, not just the technology.
What leading institutions have in common
The best initiatives aren’t defined by the technology they use; instead, they all started by asking the right questions: What is this for? Who benefits? What data do we really need? How do we protect students? How do we know it’s working?
That’s why every AI solution we use must pass a simple test: it has to be safe, evidence-based, inclusive, easy to use, and compatible with other systems.
The Decision Every Institution Must Make
UNESCO’s Observatory on AI in Education put it best: “The question is not whether AI will reach schools. It already has. The question is whether educational systems have the conditions to make that presence beneficial.” However, those conditions have to be created, as they don’t appear by themselves.
Institutions that set clear rules, connect their data, redesign assessments, and put teachers at the center, not just as users of technology, but as leaders in critical thinking, won’t just survive these changes; they’ll shape the future.
AI is like a Ferrari: if used well, it can greatly boost what we can do. If used carelessly, it can cause just as much harm. We’ve seen this happen with social media. This time, we have a chance to make a better story.
Javier Scher is SVP of Technology and Head of the Education AI Studio at Globant. Mayra Botta is the Learning Manager at the same AI Studio, where she focuses on the intersection between pedagogy and emerging technologies.
