Laptop and notebooks in an Australian classroom illustrating generative AI in schools

Australia’s Generative AI in Schools Framework: What It Actually Says

ByAIERA Blog Editorial Team

Key Takeaways

  • The Australian Framework for Generative Artificial Intelligence in Schools was developed by the National AI in Schools Taskforce on behalf of all Education Ministers and released for implementation by states and territories on 1 December 2023.
  • It is guidance, not law. It sets six principles — Teaching and Learning, Human and Social Wellbeing, Transparency, Fairness, Accountability, and Privacy, Security and Safety — supported by 25 guiding statements, and it has no enforcement mechanism of its own.
  • In June 2025, Education Ministers endorsed the 2024 Framework Review, which found the framework had correctly identified existing challenges and had effectively predicted emerging risks such as AI deepfakes.
  • Implementation sits with states, territories and individual schools, so practice varies. New South Wales, for example, gives public school students from Year 5 up access to its own tool, NSWEduChat, designed to ask students questions rather than hand over answers.
  • The framework does not substantially address academic integrity or assessment reform, and it was released with no dedicated funding and no implementation support. Schools still need their own policies on permitted tools, student data, disclosure and assessment.

A National Answer to a Classroom Question

When ChatGPT became widely available, Australian schools faced the same question as schools everywhere: ban it, ignore it, or find a principled way to use it. Australia's answer was a national framework rather than a national ban. The Australian Framework for Generative Artificial Intelligence in Schools seeks to guide the responsible and ethical use of generative AI tools in ways that benefit students, schools and society, and it has become the reference point for almost every state policy, school policy and vendor conversation since.

That is why interest in generative ai in schools australia keeps rising. The framework is short, principled and deliberately non-prescriptive, which makes it easy to cite and hard to apply without further work. This article explains what the framework actually says, who issued it, what has happened to it since, what it does not do, and what schools, teachers and parents should do with it in practice.

Who Issued It, and When

In February 2023, Education Ministers agreed that responding to the risks and harnessing the opportunities for Australian schools and students arising from generative AI technologies was a national education priority, and agreed to develop an evidence-based, best-practice framework to guide schools. The framework was developed by the National AI in Schools Taskforce, which includes representatives from all jurisdictions, education sectors and the national education agencies. Those agencies are Education Services Australia, the Australian Curriculum, Assessment and Reporting Authority (ACARA), the Australian Institute for Teaching and School Leadership (AITSL) and the Australian Education Research Organisation (AERO), and the taskforce also includes the eSafety Commissioner.

The framework was developed in consultation with teachers, students, unions, industry, academics, and parent and school representative bodies from all sectors. It was released for implementation by states and territories on 1 December 2023. The Australian Government Department of Education hosts the framework, and the OECD's policy database records it as a set of six principles supported by 25 guiding statements, developed on behalf of all Education Ministers.

Two dates are worth keeping straight, because secondary sources sometimes give different ones. Some summaries describe the framework as published in October 2023, which is when Education Ministers published it, while the government's own response to the parliamentary inquiry into generative AI in education records that it was released for implementation by states and territories on 1 December 2023. The substance is the same either way: this is a late-2023 document, written in the first year of widely available generative AI, and it has been reviewed rather than replaced since.

The Six Principles, in Plain Language

The framework is built on six principles. Each is supported by guiding statements that spell out what the principle means in a school. The principles, as listed by the Department of Education and summarised by the OECD, are Teaching and Learning, Human and Social Wellbeing, Transparency, Fairness, Accountability, and Privacy, Security and Safety.

Teaching and Learning comes first, and that ordering matters. Generative AI tools are to be used to enhance teaching, school administration and student learning, not to replace teacher expertise. They should strengthen critical thinking, creativity and meaningful engagement, not diminish them. A use that produces faster output but weaker thinking fails this principle even if it is efficient.

Human and Social Wellbeing requires that AI benefit all members of the school community without causing harm. Tools must support wellbeing, expose students to diverse perspectives, and respect human rights, autonomy and dignity. This is the principle that covers the human cost of a tool — whether it isolates students, narrows what they encounter, or puts pressure on wellbeing — and not only its academic output.

Transparency requires school communities to understand how generative AI tools work, how they are used, and how decisions informed by AI may impact them. Clear disclosure is expected whenever AI shapes processes, assessments or school operations. A parent should be able to find out whether a tool was used in marking, in a report, or in a decision about their child, and a student should know when they are working with AI-generated material.

Fairness requires AI to be inclusive and accessible for students from diverse cultural, linguistic and disability backgrounds, and implementation must consider regional, rural and remote communities. The principle encompasses accessibility and inclusivity, equity and access, non-discrimination, and respect for cultural and intellectual property. A tool that works well for well-connected metropolitan students and badly for everyone else is not a fair tool, however good its demonstration looks.

Accountability keeps human responsibility where it belongs. Teachers and school leaders, not AI systems, are accountable for decisions supported by AI. Systems must be reliable, tested and open to challenge. If an AI tool produces a wrong mark, a wrong report or a wrong recommendation, the teacher or leader who relied on it answers for it. The framework does not allow accountability to be delegated to a vendor or a model.

Privacy, Security and Safety requires AI tools to respect privacy and comply with Australian law. Schools must proactively inform communities about data collection, protect student inputs, and enforce cybersecurity measures that safeguard infrastructure and learning systems. The framework's own statements cover privacy and data protection, privacy disclosure, protection of students' inputs, and cybersecurity and resilience. This is the principle that does the most practical work in a school policy, because it is the one that decides what student data may be entered into a tool at all.

Together, the six principles and their 25 guiding statements are designed to be aspirational, guiding schools toward safe, ethical and responsible use without mandating specific implementations. That is both their strength and their limitation, and it explains most of the debate that has followed.

What the Framework Is Not: Guidance, Not Law

The most important thing to understand about the framework is its legal status. It is endorsed by Education Ministers, but it is not enforceable. There is no equivalent of a regulator conducting audits, no accountability mechanism, and no central or state body ensuring that schools actually follow the guidance. It was released with no dedicated funding, though it was related to the National Teacher Workforce Action Plan and workload reduction, and with no implementation support and no enforcement mechanism. That assessment comes from analysis of the framework's first two years, and it is consistent with the framework's own character: it offers principles and guiding statements but no roadmap for implementation.

That does not make it useless. A shared national language matters, especially when every state, sector and school would otherwise write its own rules from scratch. But it does mean that a school cannot comply with the framework simply by citing it. The framework tells a school what good looks like. The school, its system and its state still have to build the policy, choose the tools, train the staff and enforce the rules.

The contrast with higher education is instructive. In early 2025, the Australian Framework for Artificial Intelligence in Higher Education was published by the Australian Centre for Student Equity and Success in collaboration with the National AI in Schools Taskforce. It retains some of the schools framework's structure but expands it to seven principles, and it has regulatory support that schools do not have: the Tertiary Education Quality and Standards Agency (TEQSA) is a regulator with the power to enforce standards, conduct reviews and hold institutions accountable, and it has published substantial assessment reform guides that the higher education framework explicitly references. Schools have no TEQSA. That gap is the single biggest structural weakness in the schools framework, and it is worth naming plainly rather than glossing over.

The 2024 Review: Endorsed in June 2025

Education Ministers agreed that a review of the framework should take place annually, to keep up with the rapidly evolving nature of generative AI. The first of those reviews, the 2024 Framework Review, was undertaken by the National AI in Schools Taskforce in consultation with representatives of all jurisdictions, school sectors and national agencies, and was endorsed by Education Ministers in June 2025.

The review found that the framework not only correctly identified and made provision for existing challenges but also effectively predicted emerging and current risks, such as the increased use of AI deepfakes. The taskforce concluded that it was satisfied the framework provides an evidence-informed, best practice framework for the use of generative AI in Australian schools. That is a genuine vote of confidence, and it is worth noting what it does and does not claim. It claims the principles were well chosen and forward-looking. It does not claim that implementation is complete, that every school is applying the framework, or that the framework alone is sufficient.

The taskforce's work continues through two working groups, AI in Practice and AI Compliance and Standards, which deliver and oversee national work in six focus areas: workload, data security and privacy, equity, market power, research, and copyright. The AI in Practice working group, co-led by AITSL and AERO, has nationally funded projects and initiatives in the taskforce's workplan. Those focus areas are a useful map of where the hard problems sit. Workload is about whether AI actually saves teachers time or creates new work. Data security and privacy is about student data. Equity is about who gets access. Market power is about dependence on a small number of vendors. Research is about the thin evidence base for classroom benefits. Copyright is about what AI tools were trained on and what they reproduce.

What Has Happened Since: Inquiries, Plans and New Guidance

The framework has not stood still since its review. The House of Representatives Standing Committee on Employment, Education and Training conducted an inquiry into the use of generative AI in the Australian education system, reported as "Study Buddy or Influencer", and the Australian Government's response to that report was tabled in April 2026. The government response records the framework's history and the taskforce's role, and confirms that the government is providing national leadership and guidance on the use of AI in schools, collaborating with states and territories and monitoring progress on generative AI pilots in schools through the taskforce.

The National AI Plan, released on 2 December 2025, positions education as a national priority and explicitly references the schools framework. That matters because the plan also set the direction for AI regulation nationally, choosing to build on existing laws and sector regulators rather than introducing mandatory guardrails for high-risk AI. For schools, that reinforces the framework's character: national guidance, implemented through existing education systems, rather than a new AI statute for classrooms.

The evidence base is also growing, and it is more cautious than the hype. The Australian Education Research Organisation published a research report in July 2026, "Ready or not: How are schools responding to Artificial Intelligence?", which reported that around two-thirds of Australian teachers report having used AI in their work, and which examined how national principles can be combined with state-level tools that support privacy and curriculum integration. In October 2026, AERO published companion explainers for teachers and school leaders, and for policymakers, under the title "AI in schools: Considerations for school leaders and teachers". Those documents are provisional considerations, not binding national rules, and they urge schools to judge classroom AI by whether it strengthens learning, not simply by how quickly it produces an answer.

Student use is already widespread. The eSafety Commissioner reported in March 2026 that 79 per cent of children aged 10 to 17 had used an AI assistant or AI companion. That figure, reported in a 2026 guide to AI literacy in schools, underlines why a framework aimed only at teachers would be incomplete: students are already using these tools, with or without school permission, and the framework's principles on wellbeing, transparency and privacy apply to their use as much as to a teacher's lesson planning.

Internationally, the conversation is moving in the same direction. The OECD and the European Commission published an AI Literacy Framework in June 2026, organising AI literacy into four domains, and it will inform the first PISA assessment of media and AI literacy, scheduled for 2029. Australia's framework predates that work, but the two point the same way: from banning tools, towards teaching students to understand and use them well.

How States and Schools Are Actually Implementing It

Because the framework is guidance, implementation varies by jurisdiction, sector and school. That variation is the reality a parent or teacher encounters, and it is worth describing with concrete examples rather than generalities.

New South Wales has gone furthest in providing its own tool. NSW gives public school students from Year 5 up access to NSWEduChat, a tool designed to ask students questions rather than hand over answers. That design choice is a direct expression of the Teaching and Learning principle: the tool is built to preserve student thinking rather than replace it. Other jurisdictions have taken different routes, rolling implementation into their own policies and trials, and the framework's national role is to give those different routes a shared set of principles rather than a single mandated product.

State-level policies also make the practical rules explicit in ways the national framework does not. Victoria's departmental policy, as summarised in recent guides to AI in Australian education, requires opt-in consent where tools need more than a school email, sets strict limits on uploading names, reports or attendance records, and directs schools to avoid using AI to replace teacher judgement or to contact students directly. Those are the kinds of rules a national framework cannot write for every school, and they are the kinds of rules a school policy must contain if the framework's principles are to mean anything on the ground.

Curriculum work runs alongside the framework. ACARA has incorporated AI concepts into learning from Foundation to Year 10, with dedicated content in maths and technology, and science, and links to cross-curriculum priorities. The framework governs how tools are used; the curriculum governs what students are taught about AI itself. A school that does one without the other is doing half the job.

For a sense of how fast classroom AI is changing beyond policy documents, aiera.blog's coverage of Exciting Innovations: Latest Trends in AI for Education 2026 surveys the tools and trends reaching classrooms now. The framework is the lens for judging those trends, not a substitute for understanding them.

The Gaps Schools Still Have to Close Themselves

Three gaps deserve honest attention, because they are the places where a school that follows the framework literally can still go wrong.

The first is academic integrity and assessment. The schools framework does not substantially address academic integrity and the need for assessment reform, instead directing readers to assessment reform guidance published by TEQSA, including "Assessment Reform for the Age of Artificial Intelligence" and "Enacting Assessment Reform in a Time of Artificial Intelligence". That is a real limitation. The question most teachers actually face — how do I set work that AI cannot do for the student, and how do I know whose work I am marking — is not answered by six principles. Schools need assessment policies that assume AI is available, that design tasks around process and in-class demonstration rather than take-home output alone, and that are explicit with students about what AI use is permitted, what must be disclosed, and what counts as misconduct.

The second gap is resourcing and practical support. The framework was released with no dedicated funding and no implementation support, and analysis of its first two years notes that it has been criticised for a lack of resourcing and practical support. A small school with no IT specialist, no policy officer and no training budget cannot turn six principles into a working policy by reading a seven-page document. That school needs templates, vendor assessments, training time and someone to answer the question of whether a particular tool may be used with Year 7 students. The taskforce's working groups and AERO's explainers help, but they do not close the gap on their own. School leaders should budget time and, where possible, money for this work, and should treat it as an ongoing responsibility rather than a one-off policy launch.

The third gap is evidence. AERO's 2026 guidance urges schools to judge AI by whether it strengthens learning, and that is the right test, but the evidence on classroom benefits and risks remains thin. A tool that produces polished essays, faster marking or tidier lesson plans has not necessarily improved learning. Schools should pilot tools with a clear question in mind — what learning outcome should improve, and how will we know? — and should be ready to stop using a tool if the answer is disappointing. The framework's principles support that discipline, but they do not enforce it.

What a Good School AI Policy Looks Like Under the Framework

A school that wants to put the framework into practice does not need to invent its approach from scratch. The following elements map directly onto the six principles, and a policy that covers all of them is applying the framework rather than merely citing it.

Start with permitted tools. List which generative AI tools are approved for staff use and which, if any, are approved for student use, and say what happens if someone uses an unapproved tool. A list that is reviewed regularly is worth more than a blanket statement that AI must be used responsibly. For school leaders weighing dedicated tools, SLT AI: What It Is, Features, Pricing, and How It Helps School Leaders is a useful example of the kind of product a school needs to assess against the framework before adopting it.

Then set the data rules, and make them specific. No identifiable student data — names, reports, attendance records, assessment results, health information — should be entered into a consumer-grade AI tool. Where a tool requires more than a school email address, require opt-in consent. Ask vendors, in writing, what data they collect, where it is stored, whether it is used to train models, how long it is retained, and how it can be deleted. If a vendor cannot answer those questions, the tool does not meet the Privacy, Security and Safety principle, and the school should not use it. Privacy law already requires schools to handle personal data with integrity, transparency and accountability under the Australian Privacy Principles in the Privacy Act 1988 (Cth), and the framework's privacy principle sits on top of that legal duty rather than replacing it.

Set the disclosure rules. Teachers should disclose when AI has been used to produce or substantially shape material given to students or parents, including reports and feedback. Students should disclose when AI has been used in their work, and in what way. Disclosure is not a punishment; it is the Transparency principle in action, and it is what allows a teacher to judge a student's work fairly.

Set the assessment rules. Decide, subject by subject, what AI use is permitted in each type of task, and write it into the task instructions rather than leaving it to be inferred. Design at least some assessment that AI cannot complete on a student's behalf, such as in-class work, oral explanation, or work that builds on classroom discussion. The framework's silence on assessment reform makes this the school's responsibility, not the framework's.

Protect teacher judgement explicitly. State that AI does not replace teacher judgement, that no AI-generated mark, report or recommendation is final until a teacher has reviewed it, and that teachers remain accountable for decisions supported by AI. That is the Accountability principle, written as a rule rather than a slogan. It also protects teachers: a teacher who can show they reviewed, checked and took responsibility for an AI-assisted decision is in a far stronger position than one who passed a tool's output straight through.

Finally, cover equity and access. Check whether a tool works for students with disability, students from diverse cultural and linguistic backgrounds, and students in regional, rural and remote communities with limited connectivity. If a tool is only available to students who can pay for it, or only works well for some students, the school needs a plan for the students it leaves behind. The Fairness principle requires that check to happen before adoption, not after a complaint.

Parents have a role in this too. A parent should be able to ask a school four questions and get a clear answer: which AI tools does the school use, what student data goes into them, how is AI use disclosed in assessment, and who is accountable when an AI-assisted decision is wrong. If a school cannot answer those questions, the framework is not yet implemented there, whatever the school's website says.

Why This Framework Still Matters in 2026

It would be easy to dismiss a seven-page, non-binding, unfunded framework as a symbolic document. That would be a mistake, for two reasons. First, the framework's principles were well chosen. The 2024 review found that it had correctly identified existing challenges and effectively predicted emerging risks such as AI deepfakes, which is a stronger record than most technology policy documents manage. Second, the framework has become the common language for a system that would otherwise fragment. States, sectors, vendors and schools all refer to the same six principles, even when they implement them differently, and that shared language makes comparison, challenge and improvement possible.

The framework's future depends on the annual review cycle that Education Ministers agreed to, and on whether the next reviews are accompanied by the resourcing and practical support that the first version lacked. The 2025 Framework Review, and the work of the taskforce's working groups on workload, data security and privacy, equity, market power, research and copyright, will show whether the framework evolves from principles into practice. The National AI Plan's explicit reference to the schools framework suggests it will remain the national reference point, and AERO's 2026 guidance suggests the evidence base for judging classroom AI is finally being built.

For teachers, the practical message is encouraging rather than restrictive. The framework does not ban generative AI, and it does not ask teachers to become AI experts overnight. It asks them to keep teaching and learning first, to protect wellbeing, to be transparent, to be fair, to stay accountable, and to protect student privacy. Those are things good teachers already do. The framework's job is to make sure AI does not quietly undo them.

For students, the message is similar. Learning to use AI well — to check its output, to disclose its use, to think critically about what it produces — is now part of learning itself, and the curriculum work ACARA is doing from Foundation to Year 10 reflects that. The Humanize AI guide on making AI writing sound more natural is a reminder of how easily AI output can pass for a student's own work, and why disclosure and assessment design matter more than detection alone.

Australia's framework for generative ai in schools will not settle the debate about AI in classrooms. No seven-page document could. But it gives every school in the country the same six questions to ask before adopting a tool, setting a task, or entering a student's data into a system. A school that asks those questions honestly, answers them in a written policy, and revisits them as the technology changes is doing what the framework asks. A school that cites the framework and changes nothing is not.

Sources

Similar Posts