Megan Gilmour’s opinion piece was published in The Mandarin on 28 July 2026. To view the original article click here.
Australia’s AI strategy overlooks the children who will make up its future workforce, risking capability, safety, and economic success.
This month, Prime Minister Anthony Albanese announced an Office of AI inside the Department of the Prime Minister and Cabinet and, with it, Australia’s first national AI framework.
The promise is investor clarity, faster approvals, and competitiveness, with dedicated lanes for copyright, energy, productivity, education, labour rights, and a digital duty of care.
One framework instead of a scatter of departmental positions is the right architecture. A nationally coordinated approach is long overdue, and I welcome it.
Then I listened for any mention of children and young people. They appear as targets of AI-generated propaganda, with a brief nod to chatbots, and AI in schools receives only a passing mention. So far, the framework treats children as a risk to shield, not as users to prepare.
The government’s own agencies have already measured what that framing misses. In March, the eSafety commissioner’s national survey found that nearly eight in 10 children aged 10 to 17 had used an AI assistant or companion. One in eight of those young users reported that the AI displayed controlling behaviour.
In May, ACARA’s national assessment found that most Year 10 students generate written content with AI at least monthly, while half of Year 6 students have never learnt at school how to judge whether AI-generated output is accurate. Mass adoption is behind us. School, as we know it, has already had its last day. The real question is whether children will meet these systems capable and protected, or whether they will be left to navigate them on their own.
Set those dates against the record and a pattern emerges. In June 2025, education ministers endorsed a review that found their 2023 framework for generative AI in schools to be sound. Nine months later, eSafety reported the harms. Two months after that, ACARA documented the instruction gap.
No rapid mechanism connects what regulators keep finding with what ministers have endorsed. Even the first enforceable rules, due in December, include a Children’s Online Privacy Code that does not apply to state and territory agencies. Government schools sit squarely in that gap.
An Office of AI, answering to the Prime Minister, could become the mechanism that rapidly connects regulators’ evidence with education ministers’ principles and, ultimately, policy. That is what makes the national framework’s treatment of children such a missed opportunity, and the timing makes it even more striking. A week before the announcement, 17 countries launched the Coalition for Children’s Rights and Protection in the Age of AI in Geneva. Australia was not among the founding members and has not announced any intention to join.
Safeguarding is essential, and a digital duty of care belongs in any serious framework. The instinct, when reading figures like eSafety’s, is to pull children away from the technology altogether. I understand that impulse. But protection quickly becomes protectionism, and locking children out fails them twice. It leaves them unsafe when they use AI anyway, which the evidence shows they already do, and unprepared for the economy this framework promises everyone else.
The answer is mastery. We must engage children with AI in ways that build genuine capability, supported by independent safeguarding standards built into the technology itself rather than bolted on after harm occurs.
Developing those standards is the work of the SAFE Council, on which I hold one of a limited number of voting seats. SAFE is an independent initiative defining what safe AI-mediated interactions should require for children. It brings together global expertise in education, child development, psychology, AI, governance, law, and technology.
Before taking on this role, I spent almost a decade deploying learning technology for children who were most often excluded from school. That work taught me the price children pay when a system decides they can wait.
The economics point in the same direction. Australia’s first AI-native workforce is sitting in Year 10 classrooms right now. Every productivity claim in last week’s announcement is, ultimately, a claim about them, and about the Year 6 student who still cannot tell whether the machine is right.
The new office therefore has its first decision to make. Write children into the national framework as users to prepare. Give the education lane a capability mandate. Build eSafety’s and ACARA’s findings into a fixed reporting cycle. Adopt independent safeguarding standards for every AI product that reaches a child. Then put Australia’s name on the Geneva coalition.
Much of this requires no new legislation. It requires the office to decide that children are core business, and to make that decision early. An Office of AI sits at the centre of government so these decisions can be made once, for the whole country. Let’s make this one first.
