UNICEF’s September 2026 report suggests schools should teach more than prompting: students also need verification, privacy judgement, foundational learning and moderation. Because the findings come from consultations with 97 children in five African countries, schools should use them to design and test local questions rather than claim a universal curriculum.
What UNICEF Published
UNICEF Innocenti published *Skills For An AI World* in September 2026. It reports child-centred consultations with 97 participants aged 9–17 in the Democratic Republic of the Congo, Egypt, Mozambique, Senegal and Rwanda. The children discussed how they use and understand AI, what worries them and which capabilities they believe matter. This is useful recent evidence, but it is not a global survey or an effectiveness trial. Schools elsewhere should treat the findings as questions to test locally, not as a ready-made curriculum or a claim about every student.
Ask What Students Actually Mean By AI
The participants described AI mainly through chatbots, although AI also appears in recommendation and predictive systems. Older adolescents reported more sophisticated use than younger participants. Before introducing another tool, a school could ask students to map where they encounter AI, what they believe it does and what evidence supports that belief. The gap between using a chatbot and understanding wider AI systems matters. A lesson that teaches only prompting may improve one interaction without helping students recognise data collection, recommendations, automation or system limits.
Protect Foundational And Transferable Skills
Children in the consultations valued prompting, but they also highlighted critical thinking, information literacy, privacy, literacy and numeracy. They wanted to verify outputs and avoid over-reliance. That challenges a narrow “AI skills” programme built around producing faster answers. Schools can ask: which part of this task should the learner still perform unaided, where must evidence be checked, and what skill should remain visible in the final work? Your Child Uses AI. Can They Explain When It Is Wrong? offers a practical family version of that verification question.
Teach Privacy As A Decision
Participants saw safety, data protection and privacy as important, but disagreed about what information was safe to share. A school policy therefore needs examples, not simply an instruction to “protect your data”. Students could classify fictional prompts containing a name, location, health detail, image or another person’s information, then rewrite them with less disclosure. The exercise should connect classroom rules to the approved tool’s actual settings and terms. It must not imply that removing a name makes every disclosure safe.
Check Who Is Missing
UNICEF recommends grounding skills initiatives in local realities and addressing unequal access to devices, connectivity and relevant languages. A compulsory AI task can disadvantage a student who lacks reliable access or needs an unsupported language. Before assessing AI-related work, schools should identify the device, connection, accessibility and language assumptions built into it, and provide an equivalent route. They should also ask students what support they want. Consultation does not mean students alone decide safeguarding or procurement, but their experience can reveal barriers that an adult-designed checklist misses.
Ask What The School Can Defend
A school should be able to explain why a particular system is being used, what learner data it receives, how teachers will review outputs and what happens when the tool is wrong or unavailable. It should distinguish a classroom experiment from an endorsed source of truth. Staff also need time to compare practices across subjects: acceptable support in brainstorming may be unacceptable in an assessment designed to show independent reasoning. Record the decision and review date so rules can change with evidence and technology.
Run A Small Classroom Audit
Choose one existing assignment rather than launching a new AI programme. Ask four questions: What learning must the student demonstrate? Where could AI help without replacing that learning? What claims and sources must be verified? What personal or third-party information must stay out? Then test the task with and without AI, record where judgement was required, and revise the instructions. For a careful way to read youth-AI statistics, see Children And AI: Read The Evidence Behind The Headline. The immediate next step is a documented classroom test, not a prediction about future jobs.
Sources And Further Checks
Primary sources checked on 10-10-2026. Course and policy details can change; check the current official information before acting.
UNICEF Innocenti: Skills For An AI World UNICEF Innocenti: Skills For An AI World Working Paper UNICEF Innocenti: Where We Stand TodayAI-assisted production. How We Produce Stories · Report A Correction
