AI in Education: Challenges, Boundaries and the Future

It’s here. We all know it’s here to stay. AI is a reality that no one can run away from. I write this article as there is unprecedented pushback against the rollout of AI models by the very same people who created it. As the world ponders the guardrails required to protect the growth of AI (and as some would hypothesize, the future of humankind), the questions that are finally being asked have been on the mind of educators since late 2022, when ChatGPT made its way to the fingertips of consumers.

How do we protect our students from the AI onslaught that is around us? How do we ensure that cognitive development is not compromised as we begin to introduce AI use in the classroom? These are not easy questions, and there are no straightforward answers. We must carefully balance technological literacy with foundational critical thinking skills, ensuring that students retain their capacity for independent analysis, creative problem-solving, and deep focus. Bottom line – we have entered the unknown when it comes to student AI use, and right now no one really knows the true long-term cognitive and developmental impact that AI will have on students.

So what do we know about AI use? First, it holds the promise to individualize education for every student. If we get this right, every student with access to a device will be able to customize their education pathways, receive real-time adaptive feedback, and explore complex topics at their own pace in ways that were unimaginable prior to 2022. AI-driven intelligent tutoring systems can scaffold learning for struggling students while providing accelerated enrichment for advanced learners. Finally, we seem to have found the elixir we are looking for.

However, just finding the elixir is not enough. We are now almost 4 years in and this is what I observe:

1. Average OECD performance in the PISA 2025 assessments reached its lowest level since PISA began, with ongoing drops in reading and mathematics.
2. Cheating is on the rise: according to a New York Times article, a third of students have already cheated by the time they graduate from elementary (primary) school!
3. Teachers are using AI to answer emails (in many cases without reading them), make lesson plans (without putting thought into them), and mark assignments (without providing constructive feedback).

There are exceptions of brilliance, but the downsides seem to outweigh the upsides at this time. The elixir is not meeting its potential. So where do we go from here? I have the following suggestions:

1. Ban all AI use in elementary (primary) schools. I mean all – no teachers, no students, no leaders. Early childhood and primary education must prioritize neurodevelopmental milestones, foundational reading fluency, handwriting, basic numeracy, and social-emotional interactions. Over-reliance on automated tools at this age risks truncating core cognitive pathways. Keeping the sanctity of literacy and numeracy development without introducing AI ensures cognitive development continues on the basis of grounded research that has stood the test of time.
2. Introduce a structured approach to AI in middle school (lower secondary) to adapt learning to meet the student’s needs. At this stage, AI should serve strictly as a guided learning assistant rather than a shortcut. This can be done through platforms that continue to adapt challenges in core subjects like Mathematics, as well as junior-prompt engineering boot camps that teach students how logic, structure, and query precision drive computational responses.
3. In high school (secondary), clearly identify subjects where AI use will be permitted and those where it won’t. Where it is permitted, shift the focus toward high-level evaluation and synthesis. Do away with easily automated summative assessments – stick to ongoing formative assessments, oral defenses, and collaborative projects throughout the year. Adopt pedagogical approaches such as the “AI-as-learner” model, where students act as teachers and critique AI output, debug algorithmic hallucinations, and identify gaps in both the model’s output and their own understanding.
4. Have a mandatory AI ethics class in middle and high school to ensure students understand the risks of AI while promoting its effectiveness.
5. Ensure all students, when they graduate high school, know how to effectively prompt, create agents, and vibe code. These are skills that will come in handy as they move into the next phase of their lives.