AI Literacy in African Schools: Policy vs Classroom (2026)

Last reviewed: October 2026.

African education systems are writing AI into strategies, bills and curricula faster than most classrooms can absorb it. That is not a criticism. It is the right instinct. But a commitment in a national strategy and a lesson that helps a child judge an AI output are very different things, and the gap between them is where most of the work sits.

This page sets out what AI literacy should mean for school learners, what four systems have actually committed to, the gaps that keep appearing, and what ministries and schools can do next. The detailed case behind it is AI literacy in African schools: from policy ambition to classroom learning, which looks closely at Ghana’s proposed primary curriculum.

What AI literacy means for school learners

For school purposes, AI literacy is not knowing the names of tools. It has four parts that need to be taught and assessed:

Understanding. Enough grasp of how AI systems work to question their outputs: that many systems learn patterns from examples, that the examples affect the results, and that an automatic response is not the same as human understanding.
Use. Choosing a tool for a reason, and being able to explain what help it gave and what the learner still had to do.
Evaluation. Checking claims against evidence, spotting what is missing, and deciding whether an output is fit for the task.
Responsible participation. Protecting personal information, acknowledging assistance, and considering how generated content affects other people. That includes a justified decision not to use AI when a task needs independent practice.

AI literacy also overlaps heavily with media and information literacy. Recommendation systems, synthetic media and automated content already shape what learners read and watch, so evaluation skills belong in both. I wrote about that overlap in Media literacy and artificial intelligence.

Country snapshot: Ghana, Kenya, Nigeria and South Africa

These four countries are not representative of the whole continent, but their public documents show the direction of travel clearly. At continental level, the African Union’s Continental Artificial Intelligence Strategy was endorsed by the AU Executive Council at its 45th Ordinary Session in Accra in July 2024, after African ICT ministers adopted it in June 2024.^[1] It includes action on AI skills and on information integrity.

Ghana

Ghana has an adopted national plan: the National Artificial Intelligence Strategy 2025 to 2035, launched by President Mahama in Accra on 24 April 2026 and led by the Ministry of Communication, Digital Technology and Innovations.^[2] The strategy document proposes an “AI Ready Ghana” programme, AI training for youth from the first year of high school through tertiary education, updated secondary STEM and IT curricula with practical coding, AI and data ethics, data science and coding in primary education, and training courses for teachers.^[3]
The school curriculum is still at proposal stage. Revised curricula for kindergarten to Primary Six, presented in July 2026, explicitly include AI literacy, computational thinking, coding and digital citizenship, and in August 2026 the Education Ministry said they remained under review ahead of formal adoption. I look closely at the draft Computing and Citizenship Education syllabuses in the case page.

Kenya

Kenya adopted its Artificial Intelligence Strategy 2025 to 2030, launched by the Ministry of Information, Communications and the Digital Economy at the KICC, Nairobi, on 27 March 2025.^[4] Under talent development it commits to AI literacy across all demographics and to integrating AI into educational curricula.^[5]
Legislation is proposed but not yet law. The Artificial Intelligence Bill, 2026 was introduced in the Senate in April 2026 and is still before Parliament. It would require a new AI Commissioner to roll out AI literacy programmes nationwide, including through schools, ICT hubs and county partnerships.^[6]

Nigeria

Nigeria’s National Artificial Intelligence Strategy was published by the Federal Ministry of Communications, Innovation and Digital Economy in September 2025 as a living document.^[7] It is candid about schools: it notes that initiatives to introduce AI in schools exist but implementation and reach are limited, and records a 2023 federal directive to the Nigerian Educational Research and Development Council (NERDC) to integrate robotics and AI into the primary curriculum.
Beyond school, Nigeria has introduced a National Diploma curriculum in Artificial Intelligence for polytechnics, unveiled by the National Board for Technical Education in Abuja on 6 August 2026, with compulsory training for lecturers before institutions can offer it.^[8]

South Africa

South Africa’s Coding and Robotics curriculum for the Foundation to Senior Phase was adopted by the Department of Basic Education in June 2024. It is being introduced in primary schools, where there is capacity, one grade at a time from 2026, and the Foundation Phase version is designed to be taught “unplugged”, without devices.^[9] This is computational thinking and coding rather than AI literacy as such, but it is the curriculum home where AI literacy is most likely to land.
The national AI policy is back at the drafting stage. A draft National AI Policy, released for public comment by the Department of Communications and Digital Technologies in April 2026, proposed incorporating AI education, including responsible use, from foundational and basic education to tertiary level, with educator skilling and multilingual education where feasible.^[10] It was withdrawn in June 2026 after fictitious sources were found in its reference list, and the department aims to take a revised policy to Cabinet by November 2026.^[11]
Alongside policy, Lefa AI is a ten-month national sense-making initiative on AI in basic education convened by Human Studios with UNICEF South Africa and the DG Murray Trust. I was a panellist on its first webinar; my notes are in What AI means for Mzansi’s classrooms.

Common gaps between policy and classroom

Reading these documents side by side, the same gaps keep appearing.

Strategy before curriculum. Three of the four systems have an adopted national AI strategy; none yet has an adopted school curriculum for AI literacy as such. Ghana’s is the most concrete, and it is still a draft.
Ambition without time. Ghana’s own curriculum review was prompted partly by concerns about foundational learning and teacher overload. Adding AI content without deciding what it replaces risks deepening both problems.
Assessment is largely missing. Strategies say learners should be AI literate. Very few say how that would be assessed.
Teachers appear as a line item. Every document mentions teacher training. Few describe what teachers will actually need to teach and assess AI literacy.
Infrastructure and language assumptions. Policies written at national level often assume connectivity and English-language tools that many schools do not have.

The school-level version of this gap shows up in readiness data. When I ran readiness assessments with school leaders in Ghana, policy and ethics was the weakest dimension on average. The readiness framework explains how schools can measure where they stand.

Assessing AI literacy

If judgement is the goal, assessment has to make judgement visible. In the case page I describe an illustrative upper-primary activity: a teacher gives pupils a short AI-generated account of a local river alongside a trusted textbook passage. Pupils identify checkable claims, compare the sources, and explain which parts they would keep, correct or question. The teacher assesses the quality of the checking and the reasons given, not whether the pupil “used AI”.

Two features matter. First, it can be done with printed, teacher-prepared materials, so it does not depend on devices, connectivity or pupils opening personal chatbot accounts. Second, it serves existing subject goals, here reading comprehension and source evaluation, rather than competing with them. It is a proposed design, not a reported classroom trial, so it needs testing. For assessment more broadly, the AI Assessment Scale adapted for South African schools gives teachers a shared vocabulary for how much AI use a task allows. The full approach is in Assessment in the AI Era: Adapting the AI Assessment Scale for African Classrooms.

Teacher capability first

Ghana’s curriculum review recommended phased, adequately resourced teacher training, and its strategy proposes working through Ghana Education Service ICT coordinators. That is the right sequence. In my experience the practical requirement is broader than a demonstration of popular tools. Teachers need:

examples linked to curriculum outcomes, not generic prompts;
assessment criteria they can use with a class;
clear guidance on pupil data and which tools are appropriate;
resources that work at different levels of connectivity and can be adapted locally;
time to try things before they are expected to teach them.

Teacher capability is also where readiness data suggests the quickest wins are. For programme design, see Designing a 24-week AI training programme and the Future Educator case study. The full approach is in How to Design AI Professional Development That Changes Practice.

Language and local content

Most AI models are trained largely on English text produced elsewhere. For many African learners the language of the tool is not the language of the home or, in the early years, of instruction. Ghana’s strategy launch emphasised systems that reflect Ghanaian languages and cultural contexts; South Africa’s withdrawn draft referred to multilingual education where feasible. Both point the same way: AI literacy materials need to be examined for whose language, knowledge and experience they represent.

Infrastructure is the other half of local fit. I once watched a team offer e-books to a school with one computer in the admin office, when every teacher had a phone and used WhatsApp daily. AI risks repeating that mistake at larger scale. I have written more on this in Generative AI and education in the Global South and Adapting AI to tackle education inequality in Africa.

What ministries can do next

Define AI literacy for school purposes in a short statement covering understanding, use, evaluation and responsible participation, and publish it.
Decide where it lives. Name the subjects that carry it, such as computing, coding and robotics, citizenship or languages, and what it replaces, rather than adding a new load.
Write the assessment before the content. If you cannot describe how a learner shows AI literacy, the curriculum is not ready.
Fund teachers first. Phase in training and materials ahead of the curriculum, with resources that work offline and in local languages.
Pilot, then publish what happened. Treat early implementation as a test, with evidence on learning, not just on adoption.
Keep policy honest about status. Mark what is proposed, adopted and implemented, and review it on a schedule.

For schools working out where to start, the AI readiness framework gives a five-minute self-assessment and a 90-day plan. For system-level conversations, I am happy to talk; see Speaking or get in touch.

For the infrastructure and data decisions behind system-wide AI, see Sovereign AI for Education Systems: A Practical Guide for Ministries.

Sources

Niall McNulty is Director of Product and Innovation, Sub-Saharan Africa, at Cambridge University Press & Assessment. This guide reflects his own reading of the public documents listed above.