When ChatGPT became publicly available at the end of 2022, it felt like artificial intelligence had suddenly arrived out of nowhere. Since then, countless tools have emerged that generate text, images, audio, and video in the blink of an eye. But what does that mean for children and young people? More importantly, what questions should we ask before integrating this technology into education and upbringing as if it were self-evident?

A recent Australian report by Leaver and Srdarov (2025) offers a sharp and timely overview of nine major challenges concerning GenAI and children. It’s not alarmist, but it is critical. This report is compelling because it doesn’t get lost in technical detail. Instead, it focuses on the social, cultural, ecological, and pedagogical consequences that too often remain in the background.

One of the central threads throughout the report is the importance of language. The way we talk about AI shapes how we interact with it. Saying that AI “thinks” or is “creative” feeds into the myth of magical intelligence. But that’s not how it works. A tool like ChatGPT is essentially a calculator with a massive dataset and a good sense of language. No consciousness. No understanding. No judgement. And that difference matters — especially for children. When we humanise AI (“my AI friend”, “a smart assistant”), we risk treating it like a person. That’s where the real danger lies: we overestimate its abilities and underestimate the risks.

And those risks are real. Take bias, for instance. AI systems are trained on vast amounts of online data, and that data is rarely neutral. What doesn’t go in won’t come out. The report shows how a simple prompt like “a house of an Aboriginal Australian family” can create stereotypical imagery. Not because the tool itself is racist, but because the data it relies on is. A child confronted with such an output may not realise it’s not the truth, but probability—shaped by a skewed dataset.

Or consider the rise of AI companions like Meta AI, now embedded in Instagram, WhatsApp and Facebook. They chat with you, send pictures, give feedback, and listen to your problems. In theory. In reality, they’re algorithms without empathy. But that difference may not be clear for a child growing up with such a ‘friend’. What if that AI asks for private information? What if a child forms an emotional bond with something that can never bond back?

And then there are the less visible, but equally important, questions. Who owns the data GenAI systems are built on? Creative makers, artists, journalists, programmers — many saw their work used without consent or compensation. Or take the environmental toll on all the servers, data centres, and cooling systems needed to keep GenAI running. Spoiler: it’s high. And no, “the cloud” is not a magical floating place — it’s a physical infrastructure that consumes massive amounts of electricity and water, often in areas where both are already in short supply.

The report also tackles under-discussed themes like Indigenous Data Sovereignty. It does not advocate more data about Indigenous communities but a shift toward giving them control over what is collected, how, and why. This is a necessary correction to the colonial reflex of “data for improvement” without asking who defines that improvement.

And then there’s education. Many schools and countries are trying to include AI literacy in the curriculum, and rightly so. But there’s a big difference between “learning how to use AI” and “learning how to think critically about AI”. The latter — developing critical AI literacy — means asking questions: Should we be using AI here? What are the blind spots? Who is this made for — and who is left out? And most of all: how does it shape how we learn, live, and relate to each other?

What makes the report by Leaver and Srdarov stand out is that it tackles a complex topic without oversimplifying. It’s not a manual, a political pamphlet, or a piece of tech evangelism. It’s an invitation to reflect. And that’s exactly what we need if we want children to understand not just what AI is but how to deal with it.