AI has already made its way into schools. Now we need to regulate it.

While many education systems are still trying to figure out how to respond to artificial intelligence, students and teachers are already using it every day, often without adequate guidelines, training, or support.

This is the starting point of the analysis published by UNESCO, titled “Supporting Education as a Common Good in the Age of AI: The Case for Deliberative Governance.” The underlying question is crucial: What does an education system need to maintain its autonomy in the face of the rapid spread of artificial intelligence? And how can schools, universities, and governments shape these tools so that they support—rather than undermine—the public purpose of education?

Policies Lagging Behind Actual Use

Around the world, students and teachers are already widely using AI tools, often even before an official policy has been adopted. Many ministries of education are developing governance strategies, but this ambition is not always accompanied by the capabilities needed to carry it out.

The problem is also uneven. The ability—and sometimes the willingness—to manage these processes varies both across countries and within individual national systems. Even where AI policies exist, they often fail to reach the people they are intended to guide. Most students and teachers use artificial intelligence with little guidance or without adequate training on how to do so critically and responsibly.

The limited evidence available supports this picture, although it comes primarily from well-funded and well-connected settings. In the United States, for example, a nationally representative survey found that by 2025, more than half of students and teachers in core subjects were using AI. However, only 45 percent of school administrators reported that their school or district had a policy on artificial intelligence, and only one-third of district officials offered training on how to use it. Furthermore, when policy guidelines do exist, they are often advisory rather than binding.

The same gap is evident in higher education. While 90 percent of the approximately 400 academics involved in the global network of UNESCO Chairs reported using AI for teaching and research, only two-thirds said their institution had developed—or was developing—guidelines on artificial intelligence. Another study conducted in 28 countries found that only 4 percent of faculty members consider the guidance provided by their university to be comprehensive.

Why Is AI Spreading So Quickly?

The rapid spread of AI is driven, first and foremost, by its ease of use and perceived usefulness. Students and teachers who have access to these tools find them practical, fast, and functional. While some governments have restricted their use, many others have adopted strategies that encourage their use.

However, there is another factor—one that is increasingly significant yet less frequently discussed—commercial interests. According to the2026 preliminary report by the United Nations’ Independent International Scientific Panel on AI, in 2025, 91 percent of the most significant AI models came from the private sector. Ownership is, in fact, highly concentrated: at certain key points in the supply chain, a single supplier controls more than 80 percent of the market. The decisions that determine how these systems operate are therefore made by a small number of private companies.

This is a profound change. “Commercial” AI is not merely a new form of private involvement in education; it alters the very nature of the tools used in learning processes.

From textbooks to content-generation systems

Commercial products of the past (such as textbooks and educational software) offered materials developed by authors, which students and teachers could work with and adapt. Critical thinking, interpretation, and professional judgment remained largely in the hands of learners and teachers.

Digital platforms had already begun to shift this balance through recommendation and search systems optimized for engagement, which can influence what content a student encounters and in what order. Commercial generative AI, however, marks an even more dramatic shift, because the system no longer merely selects content produced by humans—it generates it directly. It generates responses, drafts texts, makes assessments, and performs—in real time—some of the cognitive work through which understanding normally develops.

These tools involve risks that are still difficult to define or assess, in part due to the lack of transparency regarding training data and the still-immature state of AI measurement and safety systems.

A private technology that can be governed publicly

UNESCO, however, emphasizes a key point: even though AI is now at the heart of teaching and learning, it can still be governed by the public sector.

Although these systems are designed and regulated largely outside of schools and universities, the conditions under which they operate in educational settings remain subject to public decision-making. This is where international institutions, national governments, and educational communities can and must take action.

The public purposes of education are based on two interconnected principles: education as a human right and as a common good. Every person has the right to inclusive, equitable, and high-quality education, free from discrimination. As a common good, education is co-created, shared, and supported by society as a whole, for the benefit of all.

Education truly serves its public purposes when it operates in the public interest, in public spaces, through public participation, and under public accountability. Governing AI in education therefore requires authority, decision-making capacity, and inclusive mechanisms capable of engaging governments, schools, universities, teachers, students, families, and communities.

Opportunities, if AI Is Managed Well

When used and managed properly, AI can contribute to the public goals of education. The available evidence is still mixed, but it is no longer merely anecdotal, and it points to certain areas where these tools can have positive effects.

Artificial intelligence can expand access to learning, translate materials into many languages, and help students with disabilities participate more fully through tools that adapt content, transcribe speech, or describe images. It can also lighten the workload for teachers in lesson preparation, providing feedback, and professional development.

At the same time, if left unregulated, it can widen social inequalities, become a distraction, and undermine the development of critical thinking.

That is why this issue is not just about technology, but about the very future of educational systems. AI has already made its way into classrooms and universities. The question now is whether governments and institutions will be able to regulate it before commercial interests alone redefine what it means to teach and learn.

Topic
Related
Unequal education, unequal futures
How Students’ Learning Levels Are Changing: The 2025 PISA Tests
School should reduce inequalities. In Italy, it risks perpetuating them.
When School Ceases to Be a Path to Social Mobility
Top-Ranked Schools, Unequal Opportunities: The Two Sides of Asian Education
The Middle East: Two Schools and Two Futures
Artificial intelligence alone is not enough to bridge Africa’s educational gaps
In South America, education alone is not enough to overcome inequality