A plain-English guide for RTOs: what each principle means, exactly what to do about it, and two real ASQA case studies that show the difference between getting it right and getting it wrong.
Hereâs the uncomfortable question: if an ASQA auditor asked your team to show exactly where AI sits in your assessment process â and whoâs accountable for it â could you answer? For most RTOs, AI is already in the building. Trainers are using it. Students are using it. The only real question left is whether you can control it.
Good news first: ASQAâs new principles donât add a single new rule. They released the Principles for the Responsible Use of AI in VET (last updated 30 July 2026) to help you apply the obligations you already have when AI is in the mix. This guide gives you all five in plain English, what to actually do about each one, and two of ASQAâs own case studies so you can see it in practice.
The line most people miss: ASQA is explicit: the principles do not introduce new regulatory requirements. They give you a structured way to interpret, implement and oversee AI use within your existing obligations, including the 2025 Standards for RTOs. So this isnât more red tape â itâs a lens for the rules you already follow.
Yes. ASQA published its Principles for the Responsible Use of AI in VET on its provider guidance pages, last updated 30 July 2026. The stated aim is to guide providers towards safe, ethical and effective use of AI.
And to be clear about what they are â and arenât: the principles donât create new obligations. They give you a structured way to interpret, implement and oversee AI use within your existing obligations, including the 2025 Standards for RTOs. Treat them as the map for applying rules you already have.
Before the principles, ASQA draws a line that trips a lot of people up. These two things are not the same:
Automation follows predefined rules and does the same task the same way every time. Auto-marking a multiple-choice quiz is automation â thereâs no interpretation or decision-making beyond the rules that were programmed.
AI can interpret, analyse and generate information, and can work with complex or ambiguous inputs â giving feedback on student writing, generating lesson materials, or summarising student questions.
Why it matters: the compliance implications of a tool that interprets and generates are very different from a tool that just follows fixed rules. When you audit your own systems, sort them into these two buckets first.
ASQA is even-handed here â AI brings genuine upside and genuine risk. Notably, one of the listed risks is not engaging with AI at all. Standing still is also a choice with consequences.
| Key risks ASQA flags | Key opportunities ASQA flags |
|---|---|
| Data privacy and security | Creating administrative efficiencies |
| Algorithmic bias | Assisting with content development |
| Digital divide increasing inequality | Enhancing accessibility, inclusivity and student support |
| Decreased human interaction and overdependence on technology | Reflecting current industry standards |
| Transparency, copyright and intellectual property issues | Creating immersive learning experiences |
| Generic or inaccurate content | Enhancing performance-based assessment |
| Academic integrity risks | Tracking student progress |
| The risk of not engaging with AI at all | Making online training more engaging in real time |
Each principle below shows ASQAâs official wording, a plain-English translation, and a concrete first action for your RTO.
âAI use is supported by strong governance that ensures it does not undermine the quality or integrity of VET.â
Plain English: Someone at the top owns AI. Itâs managed, not just happening in the background.
AI canât be a free-for-all where individual trainers quietly use whatever tool they like. Your governance needs to know where AI is used, approve it, and make sure it never weakens training or assessment quality.
Do this first: Create a simple AI register â every tool in use, who approved it, and what itâs used for. You canât govern what you canât see.
âHuman oversight and accountability are maintained in all AI-supported activities, ensuring that decisions affecting students remain the responsibility of qualified trainers, assessors and staff.â
Plain English: A qualified human makes the call. âThe AI decidedâ is never an acceptable answer.
AI can assist â draft, suggest, flag â but every decision that affects a student stays with a qualified person. This is the âhuman in the loopâ idea, and itâs the single most important principle to get right in assessment.
Do this first: Name the accountable human for every AI-assisted decision point. Write it into your assessment and support procedures.
âAI systems and tools manage information securely and in accordance with existing privacy, data protection and record keeping obligations.â
Plain English: Donât leak student data into AI tools. Same privacy rules as always.
Pasting student information into a free public chatbot can breach privacy obligations in seconds. AI use has to sit inside the privacy, data protection and record-keeping rules you already follow.
Do this first: Set a hard rule â no student personal information in public AI tools â and give staff a compliant, secure alternative to use instead.
âAI use supports and enhances student equity, inclusivity, accessibility and wellbeing.â
Plain English: AI should help every learner â not just the confident, well-resourced ones.
AI can widen gaps as easily as it closes them. A tool thatâs hard to navigate can lock out students with lower digital confidence or accessibility needs. Your job is to make sure AI lifts equity, not erodes it.
Do this first: Before rolling any AI tool into learning, test it with your least digitally-confident cohort and build an alternative pathway for anyone it disadvantages.
âAI use aligns with training product requirements, industry expectations and the needs of the relevant student cohort.â
Plain English: AI output still has to match the training product, real industry practice, and your actual students.
Generic AI content can drift from unit requirements or current industry reality. Whatever AI produces still has to meet the training product and reflect what the workplace actually expects.
Do this first: Never publish AI-generated training and assessment content without a qualified person validating it against the unit requirements and current industry practice.
Unit: BSBCRT411 â Apply critical thinking to work practices
An RTO noticed a studentâs written assessment was suddenly far more advanced than their earlier work â sophisticated language and reasoning not covered in training. An AI detection tool flagged it as likely AI-generated. When asked, the student said theyâd used AI to refine their draft and improve the writing, insisted the ideas were their own, but admitted they werenât sure if that was allowed.
Because BSBCRT411 is about demonstrating thinking processes, the assessor decided written evidence alone couldnât verify authenticity. Rather than jumping straight to misconduct, they used an alternative method: a verified video response where the student explained their problem, how they evaluated information, their critical-thinking process, and their justification â submitted through an approved platform with identity verification.
The student demonstrated genuine competency verbally, though less polished than the written work â confirming theyâd leaned on AI mainly to enhance presentation. The RTO then improved its practice: clear instructions on acceptable AI use, verbal/multimodal checks, tasks requiring personal reflection, and building student AI literacy.
The lesson: AI didnât break the assessment â and misconduct wasnât the automatic answer. Combining written and verbal evidence let the RTO confirm authentic competency while supporting responsible AI use. Thatâs Principle 2 in action.
Unit: BSBMKG433 â Undertake marketing activities
An RTO introduced an AI-powered campaign tool to keep training aligned with current industry practice. At first it worked well â trainers used AI outputs as discussion starters, and students critiqued the messaging and targeting. But over time, the AI outputs quietly replaced trainer-led teaching. Students lost the chance to see how an experienced trainer interprets data and applies judgement.
When the tool was built into an assessment, some students did well â but others, especially those with lower digital confidence or accessibility needs, struggled to navigate it and couldnât demonstrate competency. The review found the problem wasnât AI itself; it was AI implemented without enough support, context or flexibility.
The fix: reposition AI as a support resource, not the main instructor. Trainers reintroduced explicit teaching, modelled how to question AI outputs, and added guided practice and feedback. The assessment was redesigned for accessibility â alternative options like analysing provided materials, templates and guided prompts â plus a short spoken explanation of each studentâs decisions to confirm authentic understanding.
The lesson: Effective AI use balances innovation with instructional support and equitable access. Technology should enhance learning, not limit it. Thatâs Principles 4 and 5 in action.
Turn the principles into something your RTO can actually do this quarter. Run every current and planned AI use through these five moves:
List every AI tool in use across admin, training and assessment, with an owner and an approved purpose.
For every AI-assisted decision affecting students, document who makes the final call.
Ban student personal information in public AI tools; provide a secure, privacy-compliant alternative.
Trial tools with your least digitally-confident learners; build alternative pathways so no one is locked out.
No AI-generated training or assessment content goes live without human validation against the unit and current industry practice.
We map your AI use against ASQAâs principles and the 2025 Standards, flag the gaps, and give you the register, policy and self-assurance evidence to close them â before ASQA asks.
Yes â published on ASQAâs provider guidance pages, last updated 30 July 2026. They donât add new rules; they help you apply your existing obligations, including the 2025 Standards, when AI is involved.
No. ASQA states the principles do not introduce new regulatory requirements. Theyâre a structured way to interpret, implement and oversee AI use within obligations you already have.
Automation follows fixed rules every time (like auto-marking multiple choice). AI interprets, analyses and generates, and handles ambiguous inputs (like feedback on student writing). ASQA says this distinction matters when weighing AIâs implications.
Yes, as long as a qualified human stays accountable for decisions affecting students and the assessment still captures the studentâs own authentic competency. ASQAâs own case study shows verifying competency through a secure video response when written evidence alone wasnât enough.
ASQAâs âResponsible use of Artificial Intelligence (AI) in VETâ and âAI case studiesâ pages under For providers â Guidance and resources on asqa.gov.au
Disclaimer:
The information presented on the VET Resources blog is for general guidance only. While we strive for accuracy, we cannot guarantee the completeness or timeliness of the information. VET Resources is not responsible for any errors or omissions, or for the results obtained from the use of this information. Always consult a professional for advice tailored to your circumstances.
Ben Thakkar is a Compliance, Training, and Business specialist in the education industry. He has held senior management roles, including General Manager, with leading Registered Training Organisations (RTOs) and Universities. With over 15 years of experience, Ben brings extensive expertise across audits, funding contracts, VET Student Loans, CRICOS, and the Standards for RTOs 2025.
By submitting this form, you agree to the VET Resources Privacy Policy.
