Four Out of Five Students Are Already Using AI at School — and Almost Nobody in Charge Has a Policy for It

The Stanford AI Index 2026 reveals a governance failure of historic proportions: the most consequential technology shift in education history is happening inside classrooms everywhere

Video summary: https://youtu.be/r-VrJk-pDQ0

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/four-out-of-five-students-are-already?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Somewhere in South Africa today, a matric student is using ChatGPT to help draft an essay. A university student is using an AI tool to prepare for an examination. A schoolchild is asking an AI chatbot to explain a concept their teacher did not have time to cover in an overcrowded classroom. None of them received any guidance from their institution on how to do this responsibly, ethically, or effectively. None of them was told what the tool can do well and what it does poorly. None of them was taught how to evaluate what AI produces, how to disclose its use, or how to ensure it is augmenting their learning rather than replacing it. They are navigating the most powerful educational technology ever created entirely on their own — because the institutions responsible for their learning have not yet produced a single coherent policy to guide them.

CONTEXT AND BACKGROUND

The Stanford Institute for Human-Centred Artificial Intelligence released its 2026 AI Index Report this month — the most comprehensive annual assessment of artificial intelligence in existence, drawing on data from across research, industry, government, and society globally. Its findings on education are among the most striking in the entire document. Four out of five American high school and college students now use AI for school-related tasks. In higher education, usage reaches approximately 90 per cent in the United States and 95% in the United Kingdom. And yet, as the Stanford HAI education chapter of the report makes explicit, only half of middle and high schools have any AI policies in place at all — and just six per cent of teachers say those policies are clear.

That gap — between near-universal student use and almost total institutional policy absence — is not a transitional problem that time will resolve. It is a governance failure happening in real time, with real consequences for the quality of learning, the integrity of assessment, and the development of the human capabilities that education is supposed to build. Students most commonly use generative AI for research, essay editing, and brainstorming — activities that are at the core of how educational systems assess learning and capability. When those activities are being shaped by AI tools that institutions have neither governed nor even discussed, the entire framework of educational accountability is operating on assumptions that no longer reflect reality.

Crucially, the Stanford Index also notes a significant finding specific to South Africa: while AI literacy has grown faster than engineering-oriented AI skills in most countries globally, South Africa is one of a small number of exceptions where engineering-focused AI skills are actually showing steeper growth than literacy skills since 2022. That inversion matters. A country producing AI engineers faster than AI-literate citizens and students is building capability in a narrow technical elite while leaving the broader population — and the education system — behind.

INSIGHT AND ANALYSIS

The international context makes South Africa’s policy vacuum more alarming, not less. China and the United Arab Emirates both mandated AI education starting with the 2025-26 school year, according to the Stanford AI Index — signalling a shift toward formal, national-level AI instruction that treats AI literacy as a foundational skill rather than an optional enrichment. As Tech Wire Asia’s detailed analysis of the global education policy landscape documents, the UAE introduced AI as a mandatory subject across all public schools from kindergarten through Grade 12, with around 1,000 specially trained teachers delivering a curriculum covering AI fundamentals, ethics, data and algorithms, and real-world applications. India is rolling out AI and Computational Thinking across all schools from Grade 3.

Singapore has integrated AI modules into primary-level computer science courses and is committed to AI training for teachers at all levels. These are not wealthy outliers pursuing innovation for its own sake. They are deliberate national decisions that the development of AI literacy is now as foundational as mathematics and reading — and that failing to provide it is a structural disadvantage that compounds over time.

South Africa has produced none of this. The Department of Basic Education has not published a national AI education policy. The Department of Higher Education and Training has not issued binding guidelines for universities and colleges. The evidence of what is happening on the ground at South African institutions is instead emerging from academic research rather than from policy. A study published in Frontiers in Education, based on in-depth interviews with 50 students across four South African universities, found that students are widely but informally using AI tools, including ChatGPT and DeepSeek for essay writing and assignment preparation — raising immediate questions about academic integrity, data privacy, and intellectual autonomy that their institutions are not yet equipped to answer. The study found that the problems are unclear AI policies, uneven digital infrastructure, and disparate staff capacity for managing AI in learning, not student unwillingness to engage responsibly.

Universities South Africa, through its Community of Practice on Digital Education, facilitated a discussion in May 2025 among representatives from North-West University, the University of the Western Cape, and the University of Johannesburg on the state of institutional AI policies. The picture that emerged was of individual institutions taking isolated steps — a draft policy under institutional review at one university, AI guidelines integrated into an academic integrity policy at another, a short learning programme for students at a third — without any national coordination, shared framework, or binding guidance to ensure consistency across the sector. Individual institutional initiative, however well-intentioned, is not a substitute for systemic governance.

IMPLICATIONS

For South African school boards, the practical implication is immediate and uncomfortable. The ETIIH EdTech News analysis of the Stanford AI Index makes a point that applies directly to every governing body in the country: adoption is accelerating, while the systems around it are still catching up. Students are using AI to complete work and support learning. Employers are testing how AI fits into workflows. Governments and institutions are beginning to define how it should be governed. The data points to a common pattern — adoption is accelerating while the systems around it are still catching up. For South African school boards, the question is whether catching up will happen proactively through deliberate governance or reactively through the discovery of a crisis.

For the Department of Basic Education and the Department of Higher Education and Training, the global comparison is a clear indictment. The countries investing most heavily in AI education at the national level — China, the UAE, India, Singapore — are not doing so because they have solved all the implementation challenges. They are doing so because they have decided that the cost of inaction is higher than the cost of imperfect action. South Africa has not yet arrived at that decision — and every month it does not, the gap between what its students are doing with AI and what its institutions know about it grows wider.

For parents and for the business leaders who will eventually employ today’s students, the implications are equally concrete. Students arriving at the workforce having used AI throughout their education without any guided framework for doing so responsibly, critically, or effectively are not the same as students who were taught to use it well. The absence of policy is not neutral. It is a choice — and its consequences will be measured in the capability, the integrity, and the AI literacy of a generation that deserved better guidance than it received.

CLOSING TAKEAWAY

The Stanford AI Index 2026 is not a document about the future. It is a document about the present — about what is happening right now, in classrooms and lecture halls across the world, including South Africa. Four out of five students are already using AI for schoolwork. Almost nobody in charge has a clear policy for it. That gap is not a preparation problem. It is a governance failure — and in South Africa, where that failure is more complete than in almost any comparable education system, and where the stakes of educational inequality are already among the highest in the world, the cost of continued inaction is not abstract. It is the capability, the confidence, and the future of a generation of young South Africans who are navigating the most consequential technological shift in the history of education entirely without guidance from the institutions that are supposed to prepare them for the world.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net