Teaching in the Dark
What Celeste Labuschagne’s Warnings Reveal About AI in South African Classrooms
eyesonsouthafrica
Amsterdam, September 12, 2026 – Walk into almost any South African classroom or lecture hall today, and you’ll likely find a strange asymmetry. The students already know ChatGPT. Many have used it to brainstorm essays, debug code, or summarise a chapter they didn’t have time to read. Their teachers and lecturers, meanwhile, are often improvising — testing tools on their own initiative, with no national policy, no institutional roadmap, and no real consensus on what “responsible use” even looks like.
This gap is exactly what Celeste Labuschagne, a PhD candidate and learning framework developer at Belgium Campus iTversity, has been sounding the alarm about. Her argument isn’t that AI is dangerous or that it should be kept out of schools — it’s that South Africa is at serious risk of repeating a mistake it has made before.
A Familiar Pattern
Labuschagne points to what happened with coding and robotics: directives were issued, curricula were drafted with the best of intentions, and then momentum simply stalled. She sees the same shape starting to form around AI. She has described her concern bluntly: the Department of Basic Education spent years building energy around Fourth Industrial Revolution skills — coding, robotics, “future-readiness” — and educators, herself included, poured real time into writing manuals and preparing classroom material, only to be left playing catch-up with no clear sense of the destination.
The deeper problem, as she frames it, is one of sequencing. There’s a genuine desire to empower teachers to adopt new technology and help learners get comfortable with it — but South Africa is putting the cart before the horse, building enthusiasm before it builds guidance. Without that guidance, teachers are left to make high-stakes decisions about assessment, integrity, and pedagogy on their own, often quietly, sometimes inconsistently from one classroom to the next.
Why the Rest of the World Got a Head Start
Part of what makes this frustrating, in Labuschagne’s view, is that South Africa isn’t inventing this problem from scratch — it’s arriving late to a conversation that other countries have already been having. Globally, she notes, the debate has moved past whether AI belongs in schools at all; the real question now is how to integrate it responsibly, ethically and effectively, and most countries entered that debate already equipped with national AI or digital strategies, many of them refreshed since 2023. South Africa, by contrast, is still working out the fundamentals.
AI Can’t Be Policed — So Learners Need to Be Taught
One of Labuschagne’s more pointed observations cuts against the instinct many institutions have to try to detect and punish AI use rather than teach around it. Because AI cannot be policed any more effectively than Google could be, she argues, learners instead need to be taught how to engage with and integrate these tools critically and honestly — otherwise the education system does them a disservice, sending them into a world where AI is embedded in everyday systems and decision-making without giving them the underlying skills to navigate it.
This reframes the anxiety many teachers feel. The question isn’t really “how do we stop students from using AI to cheat” — it’s “how do we teach students to use AI without losing the thinking that matters.”
Teachers First, Then Learners
Crucially, Labuschagne doesn’t see this as a problem that starts with students — it starts with teachers. She argues that active, technology-enabled learning requires confident, skilled educators, and that teachers who aren’t comfortable with these tools will inadvertently suppress their use, no matter how good their intentions are. Technology deployed without intention, in her words, produces little of value.
She’s also careful to note that the infrastructure argument — the idea that South Africa isn’t “ready” for this — doesn’t hold up as well as it once did. About 80% of public schools now have internet connectivity, with national penetration nearing 75%, and the devices are already in learners’ hands. The real question, she suggests, is whether schools use that fact deliberately, or keep confiscating phones at the classroom door.
The stakes of getting this wrong, in her view, aren’t abstract. A child who leaves school without the instincts to communicate with AI tools will be at a disadvantage across almost every sector of the future economy — and failing to prepare them, she warns, is setting children up for failure.
So What Does “Doing It Right” Actually Look Like?
Labuschagne’s critique is really a call for structure — for South Africa to move past pilot projects and enthusiasm and build something durable. As she puts it, the real issue isn’t whether AI belongs in education, since it’s already there; it’s whether the country can move beyond discussion and deliver the clarity, consistency and implementation that schools, teachers and learners actually need.
Drawing that thread out into practice, a few best practices emerge — for schools, universities, and individual teachers who don’t want to wait for policy to catch up:
1. Treat AI literacy as a taught skill, not a banned behaviour. Rather than policing AI use (which, as Labuschagne notes, is about as futile as policing Google searches), build explicit lessons around how to use these tools well: how to prompt effectively, how to fact-check AI output, and how to recognise when an answer is confidently wrong.
2. Redesign assessment around process, not just product. If the fear is that AI tempts students to outsource writing and creative thinking, the fix isn’t to ban laptops — it’s to make the thinking visible. Drafts, annotated outlines, in-class reflections, oral defences of written work, and version histories all reveal whether a student engaged with the material or just requested an output.
3. Separate “AI as scaffold” from “AI as substitute.” There’s a real difference between using AI to get unstuck — to test an idea, get feedback on structure, or explain a confusing concept — and using it to generate the final product wholesale. Teachers can model this distinction explicitly, showing students what productive use looks like versus what quietly hollows out the learning.
4. Invest in teacher confidence before mandating student use. Following Labuschagne’s point directly: if teachers don’t feel equipped, they’ll either avoid AI altogether or adopt it inconsistently. Professional development shouldn’t just be a one-off workshop — it needs ongoing support, shared resources, and permission to experiment without being penalised for imperfect early attempts.
5. Build institutional policy, not just classroom-level workarounds. Individual teachers making up their own rules leads to confusion for students moving between classes or institutions. Schools and universities need clear, written guidance on what’s acceptable, ideally aligned with national frameworks — even if that means institutions draft interim policies while waiting for the Department of Basic Education to catch up.
6. Use the access that already exists. With internet connectivity reaching a large majority of public schools and most learners already carrying devices, the barrier isn’t infrastructure — it’s intention. Policies that assume a “digital divide” excuse for inaction risk missing that the divide is increasingly about guidance and skill, not access.
7. Protect creative and critical thinking deliberately. The fear that AI will make students lazy is legitimate, but the antidote isn’t avoidance — it’s designing tasks that AI can’t easily shortcut: personal reflection tied to lived experience, in-class debate, hands-on problem-solving, and assignments that require synthesising multiple, sometimes contradictory, local sources.
The Bottom Line
Labuschagne’s argument isn’t a case against AI in education — it’s a case against drift. South Africa has been here before with coding and robotics: good intentions, real investment, and then silence where a plan should have been. Whether AI in schools becomes a genuine opportunity or another stalled initiative depends less on the technology itself and more on whether the country builds the scaffolding — policy, teacher training, and thoughtful assessment design — before the gap between what students are already doing and what teachers are prepared for grows any wider.
