by Bill Corcoran
Artificial intelligence (AI) is widely used in classrooms globally, but public governance is struggling to keep pace with its adoption, a new UNESCO discussion paper warns. Students and educators globally are using AI tools daily, often without any institutional guidance, training, or policy support, according to the briefing.
Used well and governed deliberately, AI can benefit everyone involved in education, the paper’s authors believe. The problem is the absence of governance, which needs to be introduced to oversee the technology as a matter of urgency. Titled Sustaining Education as a Common Good in the Age of AI: The Case for Deliberative Governance, the paper calls for urgent and inclusive public decision making to ensure AI supports education rather than undermines it.
The paper paints a worrying picture of a technology that has moved faster than the systems meant to govern it. In the US, for instance, more than half of school students and core-subject teachers were using AI by 2025, yet fewer than half school principals reported having any AI policy in place. Globally, while 90 percent of university academics surveyed by UNESCO said they use AI for teaching and research, only 4 percent felt their institution’s guidance on the matter was comprehensive.
While ease of use and government strategies play a role in driving AI’s rapid spread in education, the paper’s researchers identified commercial pressure as the most significant – but least examined – factor behind its proliferation in the sector. According to the 2026 United Nation’s preliminary report on AI, 91 percent of notable models originated in the private sector in 2025, with a single provider controlling more than 80 percent of key supply-chain steps. “The pedagogical acts through which education realizes its public purposes, how learners reason, how teachers judge, and how learning is assessed and recorded, are increasingly mediated by AI systems designed outside education,” the UNESCO discussion paper states.
One of the paper’s most sobering findings concerns “the illusion of competence”. A study of nearly 27,000 secondary students found that those who completed homework using AI scored higher and finished faster – but their exam results fell significantly within six months when AI support was taken away.
Critical thinking, the ability to monitor one’s own understanding, and human connection are identified as the capacities most at risk. Half of US students surveyed reported feeling less connected to their teacher when AI was used in class, and seven in ten teachers worried it was weakening essential skills.
The paper also highlights serious equity concerns. Some 2.2 billion people remain offline globally, and where connectivity exists, the most capable AI tools sit behind paywalls. Furthermore, most AI systems are trained on data from a narrow range of languages and regions, meaning learners in lower-income countries, or those who speak minority languages, are at a major disadvantage.
Ungoverned AI integration is reshaping education in numerous ways, the paper goes on to say. It erodes teachers’ professional judgement and treats learners as data subjects that commercial platforms can monetise without meaningful public oversight. In response, it proposes policy actions that include requiring AI systems to be auditable, prohibiting the use of learners’ data to train commercial models, and embedding critical AI literacy into every education system.
Importantly, the paper also calls for the evaluation of learners on their understanding of a subject once the AI tool they use to delve into it is taken away. What is needed, states the paper, is “deliberative governance” – public decision making about AI that is reasoned, inclusive, and gives education stakeholders a voice in the process.
