This TEQSA generative AI guide is written for the people who actually redesign units. TEQSA's resources reduce to two principles, that assessment must equip students for a society in which AI is pervasive and that trustworthy judgements about learning require multiple, inclusive and contextualised approaches, and those principles translate into decisions about tasks, secured assessment points, what students are told, and how integrity conversations are handled. None of it is a Threshold Standard, but all of it shapes how the standards on assessment and integrity are now read.
This article walks a course coordinator through what the regulator has published, what it means for a unit, and what a provider's institutional plan expects of teaching staff, drawing on fifteen years of TEQSA registration and accreditation work.
The documents behind any TEQSA generative AI guide
There are three documents worth reading in full. The first is the 2023 paper Assessment reform for the age of artificial intelligence, which sets out the two principles and propositions favouring assessment of process, secured points and program-level design. The second is the June 2024 request for information, which required every provider to lodge a credible institutional action plan, overseen by governance, by 3 July 2024. The third is the 2025 paper Enacting assessment reform in a time of artificial intelligence, which moves from principle to practice.
Two cautions apply. These publications are guidance, not Threshold Standards, so a provider is not in breach of a paper. And TEQSA has not endorsed detection software as sufficient, so a unit whose integrity strategy is a similarity score has not engaged with the guidance at all. Our overview of TEQSA's generative AI toolkit summarises the knowledge hub.
What the two principles mean for unit design
The first principle, that students must be equipped for a society pervaded by AI, means a unit cannot simply prohibit the tools. Somewhere in the course, students need to use generative AI in a disciplinary way, evaluate its output and understand its limits, and that use should be taught and assessed. The coordinator's question is which units carry that learning and how it maps to course learning outcomes under Standard 1.4.
The second principle, that trustworthy judgement requires multiple, inclusive and contextualised approaches, is the one that changes assessment design. No single task, and certainly no single unsupervised written task, can be relied on to confirm a learning outcome. The sector's working answer has been program-level assurance: identify the points where achievement must be confirmed under secure conditions, and design the rest of the assessment for learning rather than for proof. Our article on what AI means for course design compliance covers the course-level view.
Redesigning assessment tasks in practice
In my experience the useful starting point for a unit is to sort its existing tasks into three groups: tasks where AI use is expected and declared, with assessment focused on what the student did with the output; tasks where AI use is permitted within stated limits; and secured tasks, meaning invigilated, oral, practical or in-class, which carry the burden of confirming that the outcome was achieved.
The TEQSA generative AI guide I give coordinators is to make the secured points few, deliberate and mapped, so that every course learning outcome has at least one secured confirmation somewhere in the course. Process evidence matters in the other tasks: drafts, reflective commentary, version histories and short vivas make an unsupervised task more trustworthy without pretending it is secure. Rubrics should reward judgement about AI output rather than penalising permitted use.
What to tell students, and where
Standard 7.2 requires accurate information to students about what is expected of them, which here means every unit outline stating what AI use is permitted for each task, how it must be acknowledged, and what happens if the rule is broken. Vague phrases about "appropriate use" do not meet that expectation, and in my experience they are the first thing an assessor points to when a student complaint reaches TEQSA.
Consistency across a course matters as much as clarity within a unit. Students who face a different AI rule in every unit cannot be expected to comply, and the provider cannot show the coherent design the 2023 principles ask for. The institutional action plan lodged in 2024 should set the framework; the coordinator's job is to apply it and say so in the outline.
Integrity conversations and academic misconduct
Standard 5.2 requires policies and processes that address academic integrity, and generative AI has changed what a credible process looks like. A misconduct finding based only on a detection score is, in my view, indefensible on appeal. What works is a conversation: a short interview in which the student explains their work, the process evidence is reviewed, and the marker's judgement is recorded.
That conversation should be a normal part of assessment rather than an accusation. Where a unit routinely asks a sample of students to talk about their submissions, the interview is not stigmatising and integrity evidence accumulates without a case being opened. The record of those conversations, and of any decisions that followed, is what an assessor will read under Standard 5.2. Our article on assessment integrity and TEQSA best practice sets out the policy side.
The coordinator's part in the institutional plan
Every provider lodged an institutional action plan in 2024, and TEQSA said it would follow up plans that were insufficient. The weak point of most plans is not the policy but the evidence that teaching staff have acted on it. The regulator will ask, at renewal or in a compliance assessment, for redesigned unit outlines, mapped secured points, student guidance and integrity records. Those documents are produced by coordinators, not committees, and no TEQSA generative AI guide can substitute for them. If the plan says assessment has been redesigned and your unit outline has not changed since 2022, the gap is visible.
Download the AI Governance Checklist
— a one-page check for coordinators and academic boards covering task design, student guidance, integrity process and evidence, drawn from our TEQSA registration and governance work with private providers. Get the checklist
Want the full article?
Enter your email for free access to the rest of this guide and our TEQSA resource library.
Frequently asked questions
Does TEQSA prohibit students from using generative AI?
No. TEQSA's 2023 principles state that assessment should equip students for a society in which AI is pervasive, which implies taught and assessed use of the tools. What providers must do is decide where and how use is permitted, tell students clearly, and secure the assessment points that confirm learning outcomes.
Is AI detection software enough to protect assessment integrity?
No. TEQSA has not endorsed detection software as sufficient, and a misconduct process resting on a detection score alone is weak on appeal. Process evidence, secured assessment points and a conversation with the student are the elements the guidance points to.
What is a secured assessment point?
A task completed under conditions that allow the provider to verify the student's own achievement of a learning outcome, such as an invigilated examination, an oral examination, an in-class task or a supervised practical. The sector approach is to map a small number of these across a course so every outcome is confirmed at least once.
What did the 2024 request for information require?
TEQSA asked every registered provider to lodge, by 3 July 2024, a credible institutional action plan overseen by its governance bodies addressing the risk generative AI poses to award integrity, and said it would follow up plans that were absent or insufficient.
Dr Brendan Moloney is CEO of Darlo Higher Education, Australia's largest specialist TEQSA consultancy. He holds a PhD from the University of Melbourne, is a Cambridge University Press author on governance in higher education, and has advised private providers on registration and course accreditation for more than fifteen years.