https://www.esafety.gov.au/

Access via our website.

_______________________________________________________________________

Safe On Socials

After a digital safety session in Sydney last week, a Year 11 student wrote to me. He had approached me after the session with a question about my thoughts on AI use. One of those incredibly curious and deep thinking young minds. So much so that I said to the teacher near me “He will be the Prime Minister one day”.

Yesterday he emailed me through our contact page on the website because something had been bothering him, and he had more questions for me.

He said he had watched AI quietly change shape in front of him. One week it was a tool for learning. The next, it was standing in the place where learning used to happen.

He said it better than most of the adults in the room ever have. The danger, he wrote, was that AI could “very quickly become a substitute for thinking, writing, and struggling through work independently.” Give it enough time, and he worried it would hollow out the very things school is supposed to build; the thinking, the confidence, the stamina, the simple ability to manage without a machine whispering in your ear.

Please read that again, because a sixteen-year-old just diagnosed the entire debate in a single sentence.

And notice what it is not. It is not a frightened child swearing off technology. It is not nostalgia for chalk dust and fountain pens. It is a young person looking the adults in the eye and saying, gently. “Please don’t let convenience eat my capability while everyone’s busy admiring the software.”

This is exactly why student voice belongs at the front of this conversation, not stapled to the back of it.

Students are not the customers of the education system. They live inside it. They know what is really happening, in the classroom and in the group chat and at eleven at night when a deadline is breathing down their neck and a chatbot is offering to make the whole problem disappear. They know which tools help them understand, and which ones help them avoid understanding and feel clever about it. They know where the cracks are because they are standing on them.

So if we are serious about preparing them for the future, we cannot keep making decisions about technology over their heads and then handing them the consequences. They have to be in the room as witnesses, as designers, as critics, as partners.

Which means it is time to retire from a tired question. We keep asking, “Should students use AI?” Honestly, that question is beneath us now. It is too small for the moment we are in.

Here is the question worth losing sleep over. What kind of human being is this technology helping a student become?

Every piece of educational technology, AI most of all, should have to stand in front of that question and answer for itself. Not how how fast it generates content, or how convincingly it makes a tired old classroom look like the future. This is the heart of what UNESCO has been arguing that a human-centred approach, where AI serves human capability instead of quietly replacing it, with agency, inclusion, and the wellbeing of the learner kept stubbornly at the centre.

1. Because the point of education was never output. It was formation.

Students are not little content factories. They are thinkers and writers and citizens in the making, future friends and workers and leaders and parents and decision-makers. They have to learn to sit in uncertainty, to hold two hard ideas at once, to be wrong out loud, to change their mind, to defend a position and then watch someone dismantle it, and to earn the particular confidence that only ever comes from doing a difficult thing yourself and surviving it.

Technology has a place in all of that. A real one. But it has to know its place.

Its place is not the centre. The learner is the centre. The teacher is essential. The relationship, the struggle, the conversation, the slow click of something finally making sense, those are the main event. The technology is meant to stand around the edges, in service. The moment it slips from servant to substitute, we have lost something we may not get back.

An acceptable-use policy will not save us here. What schools need is a philosophy a spine of principles strong enough to hold up procurement, classroom practice, assessment, the conversation with parents, the behaviour of students, and the behaviour of the companies circling the sector with their hands out. Australia, to its credit, has begun. Its national Framework for Generative AI in Schools, signed off by education ministers, lays out principles for responsible and ethical use across the whole community, from leaders and teachers to parents, students, providers, and policymakers.

2. But a framework is the floor, not the ceiling. Here is where I think the real work begins.

Principle One. Technology should enhance learning, never replace the act of it

A good tool lets a student see further, practise harder, get sharper feedback, find help, or wander into ideas that used to be out of reach. What it must never do is the one thing that matters most and that is the thinking.

Think of the humble calculator. In the hands of a student who already understands the maths, it is freedom. In the hands of one who doesn’t, it is a very tidy way to hide that fact. AI is the same instinct with a thousand times the horsepower. It can draft, summarise, explain, translate, solve, imitate, and polish until your work gleams. Pointed well, it is a marvel. Pointed lazily, it manufactures the convincing costume of learning while the learning itself never shows up. The Education Endowment Foundation found exactly this in the evidence that technology almost never lifts learning on its own. The pedagogy around it does the lifting, and it only works when we reach for the tool on purpose, to make teaching better, rather than grabbing it because it is shiny and everyone else has one.

3. So I would ask one quiet question of every AI-assisted task, when the screen goes dark, what cognitive work is still left in the student’s hands? If the answer is “barely any,” the tool is not enhancing the learning. It has eaten it.

Principle Two. Protect the struggle that is worth protecting

Learning is not always comfortable, and thank goodness for that.

Some struggle is just damage. Confusion with no one to turn to, humiliation, overload, the slow ache of being left behind. No child should be left there, ever. But there is another kind of struggle, and it is the good stuff. The fight to wrestle a sentence into shape. To put an argument in the right order. To drag a half-remembered fact back into the light. To sit in front of a blank page long enough that your own thinking, shy thing that it is, finally walks into the room. That struggle is not an obstacle to learning. It is the learning.

The student who emailed me understood this in their bones. He was not romanticising hardship; he was pointing directly at a real cliff edge. If a machine swoops in the instant a task gets uncomfortable, a student never grows the muscle that discomfort was supposed to build. They stay smooth and untested, like a hand that has never gripped anything heavy. So we hold the line on productive struggle and treat it as precious. Not every task gets AI. Not every draft gets polished by a model. Not every question gets answered in half a second. Not every silence gets filled. Some of the most important learning of a person’s life happens in the unbearable, beautiful pause right before the answer arrives.

Principle Three. Make thinking visible again

If we want to protect independent thought, we have to be able to see it. That does not mean turning every room into a silent exam hall under fluorescent lights. It means building far more moments where students show the journey, not just hand over the souvenir at the end.

Writing done in class. Plans scratched out by hand. Ideas defended out loud, on your feet, when there is nowhere to hide. Conferences between a teacher and a student. Draft histories that show the messy middle. Reflection. Argument. Live problem-solving. The simple, revealing act of explaining why you chose that source, how your thinking shifted halfway through, what you threw away, what you got wrong and fixed.

AI has made one truth impossible to ignore, that a polished final product is no longer proof of anything. A beautiful essay tells you what was submitted, not what was understood. A flawless paragraph shows fluency, not thought. A dazzling answer might reveal nothing more than a student’s subscription. Leaning on assessment that surfaces real thinking is not an act of suspicion. It is an act of respect. We assess what is real because we take students seriously enough to want the truth about their learning.

Principle Four. Teach the difference between a hand up and a crutch

The student told me something I have not stopped thinking about. He had felt, first-hand, how easy it is to come to lean on AI, and how brutally hard it is to climb back out.

Dependency does not arrive with a bang. It creeps. First you ask it to fix a sentence. Then a paragraph. Then to sketch the plan. Then to write the draft, just this once. Then to give feedback, then to rewrite the feedback, and one ordinary Tuesday you realise you can no longer begin anything at all without it. That is not a cheating problem. That is learned helplessness wearing the costume of efficiency, and it is far more dangerous.

Students deserve to be taught, plainly, where help ends and dependency begins. They need the words for it, real examples of it, classroom norms around it, and adults calm enough to say, without panic or shame, “this use is feeding your thinking, and that one is quietly skipping it.” And when a student has already slid too far, the answer cannot only be punishment. It has to be a way back by rebuilding their stamina, their authorship, their nerve, and their trust in their own mind.

Principle Five. Treat student voice as evidence, not wallpaper

Too often, “student voice” means a nice quote on a poster or a consultation held politely after every real decision has already been made.

Students belong inside the conversations that shape them like AI policy, assessment redesign, what the school buys, the norms of the classroom, the wellbeing of the people sitting in it. Ask them what they are seeing, what pressure they are under, which tools they actually use, what genuinely helps, what quietly harms, and what the grown-ups keep missing.

The email I received is the whole argument in miniature. This student had already thought past their own interests. They were not asking, “How do I get away with it?” They were asking, “How do we stop this becoming normal for the kids coming up behind me?” That is moral leadership from someone who cannot yet vote. When a young person can name the risk of dependency, the weakness in our assessments, the case for defending your work out loud, and the maddening difficulty of getting adults to engage with any of it, the correct response is not a pat on the head. It is to build a bigger table and pull out a chair.

Principle Six. Make edTech earn its seat

Schools need to ask harder questions and accept far fewer glossy promises. Anyone who wants a place in a classroom should be able to answer, in plain human language, without hiding behind a sales deck.

What learning problem does this actually solve, and what evidence backs the claim? What data do you collect, where does it live, and who gets to see it? How is a child’s privacy protected? Does this lighten a teacher’s load or quietly add to it? Does it include the student with a disability or leave them at the margins? What assumptions are baked into the system, and what happens when it is biased, or simply wrong, or confidently hallucinating? Does the product hand agency back to the student, or slowly take it?

And then the question that tends to clear a room is what does your product ask students to stop doing?

Because every tool is a trade. Save time, lose practice. Boost engagement, kill off attention. Personalise the learning, isolate the learner from the conversation in the room. Hand over instant answers, drain away their patience. Generate gorgeous work, erode the authorship behind it. The job is not to be dazzled by what a tool can do. It is to stay clear-eyed and a little stubborn about what students still have to do for themselves.

Principle Seven. Keep the teacher at the helm

AI can write an explanation, suggest an activity, mark a simple answer, offer an example, help with a plan. Used well it can lift real weight off tired shoulders and open genuinely new doors. But let us be unsentimental about one thing it is most certainly not a teacher.

It does not know the child in front of it. It cannot read the weather of the room, the family at home, the friendship that fell apart at lunch, the confidence that has been bleeding out all term, the student smiling through a panic attack, the one nodding along while understanding nothing, the one who needs a single question asked in exactly the right way at exactly the right moment, the way only a person who knows them could ask it.

So we must refuse to shrink teachers into supervisors of software. Technology should sharpen a teacher’s judgement, not overrule it. It should offer insight, it should hand professional expertise more room, not quietly outsource it. And it should leave space for the actual art of teaching, which is human, responsive, relational, and gloriously, irreplaceably inefficient in precisely the ways that matter.

Principle Eight. Let assessment outrun avoidance

The student’s worry about assessment is the most urgent line in the whole email, and they are right. Heaps of traditional take-home tasks are now genuinely exposed, because AI can produce work that looks competent, coherent, and original enough to sail through. That does not mean every assignment is dead. It means assessment has to grow up fast, redesigned around authorship, process, and the willingness to defend what you made.

Picture a richer mix of writing under supervision, questioning out loud, planning by hand, drafting in visible stages, annotated bibliographies, reflective notes on the process, practical demonstrations, real discussion, conferences with a teacher who knows you. The goal was never to trap students. It is to make learning harder to fake and far easier to see.

And let’s talk about detection software, because it is no foundation to build a school on. A culture run on suspicion poisons the well for students and teachers alike. The braver, better answer is assessment designed so well that the evidence of real learning is simply, obviously there.

Principle Nine. Pair AI literacy with human literacy

Yes, students need to understand the machine training data, bias, hallucination, privacy, prompting, limits, copyright, attribution, the discipline of checking what it tells you. The OECD treats AI in education as a policy area that genuinely has to adapt, because these tools reshape how teaching, learning, and skills happen, and because every country now has to capture the upside without widening the divides or quietly burning down trust.4. But technical literacy is only half of it.

The other half is knowing yourself. Am I using this because it helps me learn, or because I am exhausted, anxious, bored, or quietly terrified of failing? When did help curdle into avoidance? What am I no longer practising? Could I explain this idea with the laptop closed? Do I still hear my own voice in the work, or has it been gently replaced by a stranger? Can I even remember what I wrote five minutes after I wrote it?

Those are not engineering questions. They are human ones. The future will not belong to the students who can merely use AI. It will belong to the ones who can think with it, think against it, and, when it matters most, think entirely without it.

Principle Ten. The final test is whether agency grew

In the end, the question is not whether a school used AI. The question is whether its students walk out the door more capable, more thoughtful, more articulate, more ethical, more discerning, and more sure of their own ability to learn than when they walked in. Good technology widens that. It should never narrow it.

I want every student to be able to say, and mean it “I used the tool to strengthen my thinking, but the thinking is mine. I used it to test my argument, but the judgement is mine. I used it to find support, but the effort is mine. I used it to sharpen my work, but the voice, unmistakably, is mine”.

That is the line. Cognitive sovereignty and everything worth defending sits on one side of it.

What schools could be asking righ now

Is our AI policy brave enough to face dependency, or is it still only worried about cheating? Do our assessments still show us anything true? Have we actually asked the students, or just talked about them? Have we given teachers time, training, and trust, or a login and a shrug? Are we buying this technology because it meets a real learning need, or because it makes us look like we are keeping up? And the one that should keep a principal awake, are our students leaving here more capable, or more dependent?

Above all of it, one humble instruction is to ask the students what it is actually like.

Because the young person who wrote to me has already done the very thing we claim education exists to produce. He noticed a problem and turned the mirror on his own behaviour. He thought about the people coming after them. He refused the easy comfort of convenience. And then he wrote, clearly and respectfully, to an adult they barely knew, and asked how to help change it.

That is learning. That is agency. That is the whole point, sitting in my inbox, spelled out by a teenager. One that I hope to god does become the Prime Minister.

Done right, AI can be a genuine gift to learning. It can open up explanations, offer feedback, widen perspectives, drill a skill, build a quiet confidence. But only if we stay mindful and measured. Only if we refuse, again and again, to mistake a fluent output for genuine understanding. Only if we are willing to protect the slow, awkward, deeply human work of becoming an educated person.

The future of learning was never meant to be anti-technology. It was meant to be pro-human.

And if we ever want to know whether we have the balance right, we could start in the most radical way imaginable, by listening to the students brave enough to tell us the truth.

References

1. UNESCO, Guidance for Generative AI in Education and Research (2023). Sets out a human-centred, humanistic approach in which generative AI should serve human capabilities rather than replace them, alongside guidance on regulation, data privacy, age thresholds, teacher capacity, and learner agency. unesco.org/en/articles/guidance-generative-ai-education-and-research

2. Australian Government Department of Education, Australian Framework for Generative Artificial Intelligence (AI) in Schools (endorsed by Education Ministers, 2023; reviewed 2025). Six principles and supporting guiding statements for the responsible and ethical use of generative AI across the school community — leaders, educators, support staff, providers, parents, students, and policymakers. education.gov.au/schooling/resources/australian-framework-generative-artificial-intelligence-ai-schools

3. Education Endowment Foundation, Using Digital Technology to Improve Learning: Guidance Report (2019). Concludes that technology itself is unlikely to improve learning; the pedagogy behind it is what matters, and technology works best when integrated deliberately to improve teaching and learning rather than adopted for its own sake. educationendowmentfoundation.org.uk/education-evidence/guidance-reports/digital

4. OECD, Artificial Intelligence and Education and Skills. Frames AI in education as a policy area requiring adjustment, with attention to how AI changes teaching, learning, and skills development, and to the need to capture its benefits while addressing digital divides, bias, privacy, and equity. oecd.org/en/topics/artificial-intelligence-and-education-and-skills.html

CHILD SAFETY & PARENT CODE OF CONDUCT

Link to our child Saftey and Wellbeing Policy