It's not that children are too young to grasp AI. It's that we keep framing it as an academic exercise — full of adult language, technical definitions and abstract risk.
I've seen plenty of AI resources and educators jump straight into things like synthetic media, bias, prompt engineering, em dashes and hallucinations, but then they stop short of what I think matters most for kids - exploring the impact on their lives and how it makes them feel. AI fluency is not just vocabulary; it's about judgement, confidence and self-trust.
People sometimes assume that we must be doing the same thing — another "teach kids AI vocabulary" programme. But it really isn't. After two rounds of trials, we are more confident than ever that the power is in starting with the child's world, not the expert's glossary.
Yes, concepts matter. But the average 8-year-old does not talk about bias. They do, however, have a sharp sense of fairness. They often can't spell "sycophancy"! But they do get curious about why the AI keeps agreeing with them. The word "hallucination" is very often a new one. But they have phenomenal instincts for spotting when something looks off. (Better than a lot of adults!!)
So we meet them where they are. We explore AI through their eyes. And we teach the concepts that way.