ASQA and AI: a practical guide for RTOs

ASQA's position on AI is now in writing. On 30 July 2026 the regulator released its Principles for the Responsible Use of AI in VET: five principles, each with self-assurance questions and links to the 2025 Standards for RTOs, plus two case studies. ASQA says the principles don't create new regulatory requirements. They show how your existing obligations apply when AI is involved in delivery, assessment and administration.

This guide sets out what ASQA has actually published, how the principles connect to the Outcome Standards, and the practical steps an RTO should take on training materials, assessment integrity, learner use, policy and records.

What ASQA has published about AI#

As of October 2026, ASQA's AI guidance sits in one section of its website and includes:

• Principles for the Responsible Use of AI in VET (released 30 July 2026). ASQA says these were informed by sector consultation, including its 2026 sector workshops, where it shared a draft.
• Two case studies: one on assessment integrity in a critical-thinking unit, and one on inclusive use of an AI tool in a marketing unit.
• ASQA's own AI Transparency Statement (last updated 18 March 2026), which says ASQA doesn't currently use AI in public-facing services, regulatory functions or decision-making, and that final decisions are made by a human.

ASQA also points providers to Australia's AI Ethics Principles, TEQSA's Gen AI knowledge hub and the Australian Framework for Generative AI in Schools as further reading.

ASQA draws one useful distinction: automation follows fixed rules (marking multiple-choice questions automatically is automation), while AI interprets, analyses and generates content. The principles are about the second kind.

The five principles and the Standards they map to#

ASQA's self-assurance questions are worth working through as a leadership team. Examples include whether your student or learner management system already has built-in AI features, whether staff would recognise when AI output is inaccurate, and which training and assessment practices aren't suitable for AI.

The 2025 Standards that matter most for AI#

The Outcome Standards took effect on 1 July 2025. Four parts do most of the work when AI is involved:

• Standard 1.3: assessment tools must be reviewed before use to make sure assessment can be conducted consistently with the principles of assessment and rules of evidence.
• Standard 1.4: assessors' judgements must be justified by the rules of evidence, including authenticity ("the assessor is assured that a VET student's assessment evidence is the original and genuine work of that VET student"). The validity principle also requires practical application components.
• Standard 1.5: the assessment system is validated by appropriately skilled and credentialled people, with outcomes not solely determined by whoever designed or delivered the assessment.
• Compliance Requirements clause 10: RTOs must keep records of all assessments a student submits for 2 years after they complete the training product. Clause 20 requires compliance with all applicable laws, including privacy laws.

Using AI to develop training and assessment materials#

AI is good at first drafts: session plans, scenario-based case studies, knowledge questions, marking guides, plain-English rewrites and practice activities. The risk is material that is generic, wrong or not mapped to the unit. To stay on the right side of the Standards:

• Map before you prompt. Give the AI the unit's elements, performance criteria, performance evidence, knowledge evidence and assessment conditions, and ask it to show the mapping.
• Review before use. Have a credentialled person check accuracy, currency and alignment, and record that review. That is the evidence for Standard 1.3.
• Check industry currency. AI models can be out of date. Your trainers' current industry knowledge (Standard 3.3) and your industry engagement (Standard 1.2) are the check.
• Validate as usual. AI-assisted tools go through the same validation as any other (Standard 1.5).
• Watch copyright and licensing. ASQA lists copyright and intellectual property among the risks of AI use. Don't upload licensed resources to a tool unless your licence allows it.

Assessment integrity in the age of AI#

The authenticity rule hasn't changed, but meeting it takes more deliberate design when students have AI on their phones.

Design for competency, not essays. Observation, oral questioning, workplace evidence, practical demonstrations and third-party reports are harder to outsource than written responses. The validity principle in Standard 1.4 already expects practical application components.

Don't treat a detector score as a finding. In ASQA's first case study, an AI detection tool flagged a student's written assessment for BSBCRT411 Apply critical thinking to work practices. The student said they had used AI to refine their own draft and weren't sure if that was allowed. Rather than treating it as misconduct, the assessor used an alternative method: the student recorded a video response explaining their reasoning. That showed competency. The RTO then wrote clear instructions on acceptable AI use and added verbal and reflective components to similar tasks.

Keep judgement human. ASQA lists relying solely on AI to make assessment judgements as an example of staff misuse. AI can help draft feedback; the assessor makes and justifies the decision.

Learners using AI#

Students will use AI whether or not your policy mentions it. Make the rules clear:

1. Say what's allowed for each task, for example "AI may be used to check spelling and structure, but the analysis must be your own".
2. Explain how to disclose AI use and make disclosure normal rather than risky.
3. Teach AI literacy where it reflects how the industry works (Principle 5).
4. Check equity. ASQA's second case study, on BSBMKG433 Undertake marketing activities, describes an AI tool that gradually displaced trainer-led teaching and disadvantaged students with lower digital confidence or accessibility needs. The RTO repositioned the tool as a support, reintroduced explicit teaching and offered alternative ways to complete the assessment.

Writing your RTO's AI policy#

ASQA's principles don't prescribe a policy, but they point to what one should cover. A practical structure:

• Scope and accountability: who owns AI governance, and how it reports to governing persons
• Approved tools register: each tool, its purpose, where data is stored and processed, whether inputs are used for training, and who can use it
• Information rules: what staff may and may not enter, especially student personal information
• Student consent and notice: how students are told, and give prior written consent, when their information goes into an AI system
• Staff uses: permitted uses (drafting materials, administration) and prohibited uses (making assessment judgements, entering student data into unapproved tools)
• Review requirement: AI-assisted training and assessment material is reviewed by a qualified person before use
• Student rules: acceptable use and disclosure by task, and how suspected misuse is handled
• Third parties: how partners delivering on your behalf meet the same rules (Standard 4.2)
• Training: what staff need to know to apply the policy
• Monitoring and review: when the policy and tools are reviewed

Record keeping for AI use#

Good records let you show an auditor how AI fits within your obligations. Keep:

• your AI policy and its version history
• the approved tools register and the due diligence behind it
• review records for AI-assisted training and assessment materials
• assessment evidence, including any supplementary evidence used to confirm authenticity (for at least 2 years after completion)
• student AI declarations and any consent records
• records of AI-related incidents, complaints and policy breaches

ASQA's third principle also asks: if your AI systems create records, how will you retain them as your record-keeping obligations require?

Privacy and student data#

Clause 20 of the Compliance Requirements requires RTOs to handle personal information in line with applicable privacy laws. The OAIC recommends against entering personal information, particularly sensitive information, into publicly available generative AI tools, and says privacy obligations apply both to what you put into an AI system and to output that contains personal information.

In practice, that means giving staff an approved tool rather than leaving them to use personal accounts, and checking where it stores and processes data. AusGPT is one option: it runs Claude through Amazon Bedrock with processing in AWS Sydney and Melbourne, stores conversations and documents in Australia, and doesn't use your data to train AI models. Shared prompt libraries also make it easy to give trainers consistent, approved prompts for drafting materials. See how education and training providers use it.

What to do this quarter#

1. Read ASQA's five principles and work through the self-assurance questions with your leadership team.
2. List every AI tool in use, including AI features in your student management system.
3. Approve a small set of tools and publish clear rules for staff and students.
4. Review assessment tools for units where written evidence could easily be produced by AI.
5. Add AI use to your validation and internal audit schedule.