In September 2023 UNESCO published its first global guidance on generative artificial intelligence in education and research. It arrived only months after tools capable of generating text, images and code reached mass audiences, leaving schools and universities to rethink assignments, assessment, privacy and the reliability of machine-generated answers.
UNESCO’s approach is explicitly human-centred: AI should be designed and used to strengthen human capabilities and agency rather than automatically replace the judgement of learners, teachers and researchers.
Privacy before adoption
Data protection is a central concern. Many generative tools process user prompts and may be offered under complex commercial terms. In education this matters particularly because users are often minors.
UNESCO calls for regulatory frameworks that protect privacy and institutions capable of validating tools before embedding them in teaching and learning.
Age and independent use
The guidance proposes appropriate age limits for independent conversations with GenAI platforms. A system designed for adults should not automatically be treated as suitable for children or teenagers.
The issue is not only harmful content but the ability to understand errors, bias and the probabilistic nature of model outputs. A fluent answer may still be wrong, so students need critical skills to evaluate it.
Assessment and assignments change
If a model can draft an essay or solve an exercise, traditional assessments may become less effective at showing what a student can do independently. The answer is not necessarily to ban AI, but to design activities that make reasoning processes visible.
Oral presentations, iterative work, discussion and disclosure of tool use may become more important. AI therefore forces education systems to reconsider what they actually want assessment to measure.
Pedagogical opportunities
The guidance also recognises potential benefits, including lesson design, tutoring, example generation, translation and research assistance. Their value depends on context and human supervision.
A model can help a teacher prepare differentiated material or a student explore a topic, but it does not guarantee source accuracy or neutrality. AI literacy therefore becomes part of digital literacy.
Governance before enthusiasm
UNESCO noted that platform development was outpacing national regulation. That gap remains one of the central challenges of generative AI.
Schools and universities cannot leave every decision to individual classrooms. They need clear institutional policies on data, attribution, assessment, approved tools and responsibility. AI may become an important education technology, but its value will depend on the quality of the rules and skills built around it.
