Much of Australia’s artificial intelligence conversation has played out at the top of the pyramid. It has been about sovereign data centres, the billions flowing into computing infrastructure, university research partnerships and the scramble for graduates who can build and govern models. A quieter argument is now pushing back on where the country should actually be starting, and it lands a long way from any server hall: the primary school classroom.
That is the case being made by Tom Ashmor, whose foundation has been arguing that Australia’s AI ambitions will stall unless children are introduced to the technology from their earliest years of schooling rather than being handed it as an afterthought in senior high school or first-year university. The pitch, reported by SMBtech, is that the country keeps treating AI capability as something to be bolted on late, when it should be treated as a foundational literacy laid down early.
The argument
The foundation’s position rests on a simple observation about how skills actually form. Reading, numeracy and basic scientific reasoning are not left until Year 11 because everyone accepts that the habits of thought behind them take years to build. The claim being put forward is that fluency with AI belongs in the same category. If children grow up understanding what these systems are, how they generate answers, where they get things wrong and why they cannot always be trusted, the argument runs, then the country ends up with a workforce that is comfortable with the technology rather than intimidated by it.
There is a defensive edge to the case as well. Australian children are already using generative AI, often without much guidance, to help with homework, to make images and to answer questions they would once have typed into a search engine. On that view, the choice is not whether primary-aged students encounter AI, because they plainly already do, but whether the school system shapes how they encounter it. Teaching the technology early becomes less about producing a generation of engineers and more about producing a generation that can tell a confident machine answer from a correct one.
The economic framing is familiar to anyone who has followed the national debate. Australia has spent much of the past year worrying aloud about whether it is building AI or merely hosting it, a tension that has run through everything from the federal government’s new AI office to arguments over sovereign capability. The foundation’s contribution is to point out that the pipeline of people who might one day build anything at all begins in classrooms that are, for the most part, not yet set up to introduce the subject in any structured way.
The other side
Not everyone is convinced that the answer to a crowded curriculum is to add another demand to it. Teachers and their unions have spent years warning about workload, and the practical reality is that primary educators are already stretched across literacy, numeracy, wellbeing and a growing list of digital responsibilities. Asking them to teach a fast-moving and contested technology, without significant training and clear materials, risks producing lessons that are either superficial or quietly abandoned.
There is also a genuine pedagogical debate about what “teaching AI” to a seven-year-old even means. Critics of an early push argue that young children benefit more from unplugged thinking, from building the reasoning and creativity that AI cannot replicate, than from early exposure to tools that do the thinking for them. On this reading, the risk is not that children learn too little about AI but that they lean on it too early, before they have developed the judgement needed to question it. Concerns about screen time, data privacy and the commercial interests of the companies supplying school technology sit close behind.
Those two positions are not as far apart as they first appear. Much of the disagreement is really about content rather than principle. Almost no one argues that children should be shielded from any mention of AI until adulthood. The harder questions are what an age-appropriate curriculum looks like, who writes it, how teachers are supported to deliver it and how a public school system avoids becoming dependent on the very vendors whose products it is trying to teach students to interrogate.
What it means for Australia
The stakes here are national rather than local, and they connect directly to arguments the country has been having all year about whether its AI strategy has any foundations under it. Australia has committed serious money and political attention to the infrastructure layer, backing data centres, courting chipmakers and standing up new policy machinery. The recurring criticism has been that skills and adoption lag well behind the hardware, and that a shortage of capable, confident workers could hollow out the payoff from all that investment.
An early-education push speaks to that gap in a way that infrastructure spending cannot. Data centres can be built in a couple of years. A workforce that is genuinely comfortable designing, governing and questioning AI takes a decade or more to grow, which means the children who will define Australia’s AI capability in the late 2030s are in primary school right now. If the argument is right, the decisions that matter most for the country’s long-term position are being made not in cabinet or in boardrooms but in curriculum meetings and staff rooms.
There is a fairness dimension too. If AI literacy becomes a marker of opportunity, then leaving it to chance risks widening the gap between students at well-resourced schools, where enthusiastic teachers and good devices make early exposure easy, and those in under-funded or regional schools where it does not happen at all. Building AI understanding into the early curriculum, done properly, is one of the few levers that could spread that capability evenly rather than concentrating it.
What’s next
Turning advocacy into practice is where the difficulty starts. Curriculum in Australia is a shared responsibility between the states and the Commonwealth, which makes any nationwide change slow and negotiated. Federal and state education authorities have already begun issuing guidance on how generative AI should be used in schools, but that guidance has largely concerned rules of use for older students rather than a structured plan to teach the subject from the early years.
For the foundation, the immediate task is less about grand policy and more about proof. Pilot programs, teacher training and clear, age-appropriate materials are the sort of groundwork that can demonstrate whether early AI education works in a real classroom or simply adds to the load. Whether governments pick that work up will say a lot about how seriously Australia treats the idea that its AI future depends on people it has not yet trained, sitting in classrooms it has not yet equipped.
Sources: SMBtech
