AI course design TEQSA compliance comes down to one question: can the provider show that its course still produces graduates who have the learning outcomes it certifies, now that generative AI can complete most unsupervised written tasks? TEQSA has not written a new standard for artificial intelligence. It reads Standards 1.4, 3.1 and 5.2 of the existing Threshold Standards in the light of what the technology can do, and an accreditation application that ignores that reading will draw a request for further information.
This article explains what has changed in the evidence assessors expect for course design, from learning outcomes to unit outlines, and what an accreditation submission should now contain. It draws on our TEQSA accreditation work with private providers since the regulator's 2023 assessment reform paper and its 2024 request for information.
What has TEQSA actually said about AI and course design?
Three documents matter. The first is the 2023 paper Assessment reform for the age of artificial intelligence, which set out two principles: that assessment should equip students for a society pervaded by AI, and that trustworthy judgments about learning require multiple, inclusive and contextualised approaches. Its propositions favour assessing process as well as product, securing particular assessment points, and designing assurance at the level of the whole program rather than the individual unit.
The second is the June 2024 request for information, which required every registered provider to lodge, by 3 July 2024, a credible institutional action plan overseen by its governing bodies to address the risk generative AI poses to award integrity. The third is the 2025 follow-up, Enacting assessment reform in a time of artificial intelligence, which moves from principle to practice. None of these is a Threshold Standard. They tell you how TEQSA reads the standards it already has, and in my experience assessors now read them exactly that way.
Learning outcomes: AI capability is now part of the graduate
The first change is in the learning outcomes themselves. Standard 1.4.1 requires learning outcomes consistent with the AQF level and field of the qualification, and Standard 3.1.2 requires the content and learning activities to engage with advanced knowledge and inquiry appropriate to that level. In 2026 an assessor reading a bachelor degree in business, nursing informatics or design will ask where the course develops the graduate's ability to use, evaluate and take responsibility for AI-generated work in that discipline.
That does not mean bolting a generic "digital literacy" outcome onto every course. It means the course learning outcomes, and the unit outcomes mapped to them, say what a graduate in this field is expected to be able to do with and without these tools, and the assessment tests both. The mapping matrix in the application is where this is visible. A matrix that still maps every outcome to an unsupervised essay was drafted before the technology arrived, and assessors notice.
Secured assessment points: where the judgement is anchored
The second change is structural. TEQSA's papers, and the sector practice that has formed around them, distinguish between assessment that can reasonably assume AI use and assessment that is secured, meaning the provider can vouch that the work observed is the student's own. Secured points are typically invigilated examinations, vivas, practical demonstrations, supervised laboratory or clinical work, and in-class tasks. The sector label for this is a "two-lane" approach, though that is not TEQSA's term and I would not use it in an application.
The design task is to decide which learning outcomes must be assured at a secured point, and to place those points where they anchor the award. In my experience the credible pattern is one or more secured points in each year of a program, with the capstone or final-year assessment secured for every course learning outcome that the qualification certifies. What assessors will not accept is an assurance plan that consists of a detection tool and an academic integrity policy. TEQSA has said plainly that detection software is not sufficient on its own, and an application that relies on it has not answered the question.
Program-level assurance is the evidence TEQSA now expects
The most important shift for AI course design TEQSA compliance is the move from unit-level to program-level thinking. Standard 5.3.1 requires review of all courses to be comprehensive and to include assessment, and Standard 1.4.3 requires methods of assessment capable of confirming that all specified learning outcomes are achieved. Read together, in the light of the 2023 paper, they require the provider to show for the course as a whole, not unit by unit, how each learning outcome is confirmed at least once under conditions the provider can vouch for.
The document that shows this is a program-level assessment map: every course learning outcome across the top, every unit and its assessment tasks down the side, and a marking of which tasks are secured. That map is the single most useful artefact you can put in an accreditation application in 2026. It also feeds the material the academic board needs to approve the course with its eyes open, which connects to the governance evidence described in our article on what TEQSA looks for in a new course.
Unit outlines and the rules students are actually given
Unit outlines are where policy meets practice, and where an assessor checks that the design is real. Each outline should state, for each assessment task, whether AI use is permitted, in what way, and how it must be acknowledged. A blanket sentence copied from the academic integrity policy is not enough; the permission has to be task-specific because the task design depends on it.
The outline should also match the map. In my experience the most common inconsistency in a submission is a program-level plan that describes a secured capstone while the capstone unit outline still describes a take-home report with no viva or presentation component. Assessors cross-read these documents, and an inconsistency between the plan and the outline undermines both. The broader question of what must sit in a course document is covered in our guide to course design for TEQSA accreditation.
What an accreditation application should now include
For a new course, the application should show the academic board approving a course whose assessment design has been considered against the AI risk, with the minutes recording that consideration. It should include the program-level assessment map, unit outlines with task-specific AI rules, an academic integrity policy that addresses generative AI explicitly, and a description of the staff development that will make the design work in the classroom. Where the course is a variation of an existing one, the application should say what changed and why.
For a provider already registered, the 2024 action plan is part of the record. Assessors will ask whether the course before them is consistent with the plan the governing body approved, and whether the plan has been implemented rather than filed. Standards are written in the present tense; evidence of operation matters more than evidence of intention. Our article on assessment integrity and TEQSA covers the operational side in more detail, and the wider institutional picture is in AI considerations in higher education.
What I tell clients about AI course design TEQSA questions
I tell them that the regulator is not looking for a provider that has solved AI. It is looking for a provider that has understood the problem in its own discipline, decided where the award is anchored, and built the course so that those anchor points hold. That is a design question before it is a compliance question, and the providers who do it well treat it as one.
I also tell them that generic material drafted by the same tools the policy is meant to govern is easy to recognise. Since TEQSA moved from Confirmed Evidence Tables to self-assurance, the evidentiary bar has gone up, not down, and an AI course design TEQSA submission that reads like a template invites exactly the scrutiny it was meant to avoid. Specific, true and consistent across the map, the outlines and the minutes is the standard.
Download the Darlo AI Governance Checklist
— a one-page test of whether your course design, assessment map and academic board record answer the AI question, 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 require a specific AI policy for course accreditation?
No standard names artificial intelligence. TEQSA reads Standards 1.4, 3.1 and 5.2 in the light of its 2023 and 2025 assessment reform papers and its 2024 request for an institutional action plan, and expects an accreditation application to show how the course confirms its learning outcomes despite generative AI.
Can we ban AI in all assessment instead of redesigning the course?
A blanket ban is not credible evidence because it cannot be enforced in unsupervised tasks and it contradicts TEQSA's principle that assessment should equip students for a society pervaded by AI. Assessors expect task-specific rules and secured assessment points, not prohibition.
What is a secured assessment point?
An assessment task delivered under conditions where the provider can vouch that the work observed is the student's own, such as an invigilated examination, viva, practical demonstration or supervised clinical task. A credible design places secured points so that every course learning outcome is confirmed at least once.
Is AI detection software enough to satisfy TEQSA?
No. TEQSA has said detection tools are not sufficient on their own. They may form part of an integrity approach, but the evidence assessors want is program-level assurance, with secured assessment and academic board oversight of the design.
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.