A question of method, not fashion
At school, artificial intelligence should not be presented as a magic wand. It can help students search, rephrase, compare, practise or create, but it does not replace reasoning or a teacher's guidance.
The question is not whether young people will use AI: they already encounter it. The question is whether they can recognise what it does, what it cannot do and when a person, a book or their own memory is the better resource.
AI is not human intelligence
A generative AI produces an answer from models, data and probabilities. It can write a convincing sentence while being wrong about a fact, a source or a proof. A fluent answer is never proof that it is true.
For a student, understanding a text is not getting an answer that sounds right. It is being able to explain it, check it, connect it to the lesson and defend it without depending on a screen.
The risks we need to name
The first risk is total delegation: asking a tool to think, write and conclude on the student's behalf. It may produce a polished assignment, but leaves the student without a method when the tool is unavailable or the question changes.
We also need to discuss cheating, invented information, bias, personal data collection, dependency and identical-looking work. These risks do not disappear because an interface feels pleasant.
Why schools must explain AI
Banning without explaining mostly creates hidden use. Confident students will keep experimenting, while others will not know how to assess an answer or protect their information. Education should reduce that gap.
Explaining AI means learning how to formulate a request, credit assistance, verify an answer, identify a synthetic image, respect privacy and revise one's own reasoning.
Learning with AI, learning without it
The healthiest rule is simple: use AI after trying, not instead of trying. A student may ask for a hint after a first attempt, another explanation after rereading the lesson or extra exercises after identifying a difficulty.
Foundations still need practice without assistance: reading, memorising, calculating, writing, arguing, coding and solving problems. This autonomy protects against outages, missing connectivity and tool errors.
How many AIs are there, really?
There is no single reliable counter for the number of AIs. There are models, applications, embedded assistants, specialised tools, open models and agents. The number constantly changes and depends on what we count.
It is more useful to speak about families: text and image generation, speech and vision, prediction and classification, recommendation, translation, retrieval-augmented search and agents that chain several actions.
A practical path for young people
A responsible path can start with one question: what would I have done without the tool? Then come the objective, comparison of several answers, verification in reliable sources, personal rewriting and an oral explanation of the result.
Teaching young people about AI means giving them two skills: using it with judgement and continuing without it. That combination makes the use empowering and compatible with demanding education.