Changes in V2 of ASQA's Practice Guides
On 29 October 2026, the National Vocational Education and Training (VET) Regulator, the Australian Skills Quality Authority (ASQA) released an update to their suite of “Practice Guides”.
Practice Guides were introduced to assist providers to understand their obligations and ASQA’s expectations under the 2025 Standards for Registered Training Organisations (RTOs).
Importantly, Practice Guides are not the legislation and are not intended as a legally enforceable instrument. However, they do signal what ASQA expects to see.
When we look at all of the changes from V1 to V2 of the Practice Guides, there are several interconnected themes. AI is prominent, but it is one part of a broader movement towards more active organisational accountability.
The following breaks down the overarching themes in the additional known risk areas and self-assurance questions outlined in the V2 Guides.
What the changes are signalling
Human in the loop
Greater emphasis on human accountability and oversight is perhaps the strongest recurring theme in the changes. Additions repeatedly challenge RTOs to consider whether they have allowed technology, automation or external parties to take over decisions and functions that require appropriate human judgement.
Examples include:
AI marking of assessments without clear oversight (1.4)
Over-reliance on AI when making validation decisions (1.5)
AI-assisted verification of RPL documentation (1.7)
AI-enabled student support replacing access to human staff (2.3)
Automated moderation failing to recognise cultural or linguistic nuances (2.5)
AI-generated wellbeing advice replacing appropriate support services (2.6)
AI-assisted credential verification without robust human checks (3.2)
Governing persons relying too heavily on AI to monitor compliance (4.2)
The underlying message isn't simply that AI needs supervision. It is that RTOs cannot outsource accountability for quality outcomes to technology, automated systems or third parties.
This also extends beyond AI. The changes to third-party marketing, complaints escalation and monitoring of third-party AI use reinforce the expectation that responsibility remains with the RTO, regardless of who performs the activity.
The V2 Guides give a strong focus on the actual student experience and appear to move further towards identifying what can go wrong for students as a consequence of an RTO's operational decisions.
Consider the additions concerning:
Students not having sufficient time to develop skills and apply knowledge (1.1)
Students needing digital skills to participate in AI-enabled delivery (2.2)
Access to trainers and assessors rather than relying on automated support (2.3)
Accessibility of AI-based platforms for students requiring reasonable adjustments (2.4)
Cultural safety and the potential for AI-generated content to cause cultural harm (2.5)
Protection of sensitive wellbeing information (2.6)
Accessible appeals processes and independent review (2.8)
There is an important distinction here.
The focus is not just on whether the RTO has a policy, procedure or system in place. It is whether its decisions and practices actually protect students and enable them to participate, learn, demonstrate competence and receive appropriate support.
The student experience
The student experience is increasingly being treated as an indicator of organisational quality, not simply an administrative consideration.
Another significant theme, particularly in the self-assurance questions, is the explicit expectations around evidence, monitoring and demonstrating effectiveness.
V2 introduces questions that ask RTOs to demonstrate not just that something exists, but that it works.
For example:
How do you evaluate whether industry engagement influences decisions? (1.2)
How do post-cohort reviews demonstrate improved engagement and outcomes? (1.2)
How do you determine whether AI-assisted delivery supports learning outcomes? (1.5)
How do you ensure safeguards are effective? (2.8)
How do you monitor the quality, effectiveness and appropriateness of AI use? (4.4)
Do you retain evaluation reports, monitoring plans, action registers and change logs? (4.4)
The changes to Standard 4.4 are particularly telling. Collecting feedback is not enough. Evaluating feedback is not enough. The RTO must translate that information into improvement actions and address recurring issues.
Similarly, the new questions about engaging with ASQA proactively and monitoring performance against both the Outcome Standards and Compliance Requirements reinforce a more active approach to self-assurance.
The emphasis is shifting towards being able to demonstrate that quality systems are functioning effectively, that decisions are informed by evidence, and that identified issues result in action.
Organisational governance and risk management
AI has clearly prompted additional questions about governance, but the changes go further.
The additions introduce or reinforce expectations around:
Governing persons understanding the implications of new technologies and documenting their decisions (4.1)
Clear responsibility for approving, using and monitoring AI (4.2)
Monitoring third parties and their compliance with agreed obligations (4.2)
Appropriate risk controls and data governance (4.3)
Monitoring the impact of AI on workforce capability (3.1)
Documenting decisions relating to the suitability of governing persons (Fit and Proper Person Requirements)
Timely notification of material changes, including matters that have been developing over time (Compliance Requirement 16)
The broader implication is that governance cannot simply be about having policies and delegating responsibility. RTOs need to understand what is happening across their operations, identify emerging risks, make informed decisions and be able to demonstrate how those risks are being managed.
Explicit protection of the integrity of training, assessment and qualifications
This is particularly important in Quality Area 1 and the Compliance Requirements.
The changes address risks associated with:
Accelerated or shortened training that doesn't provide sufficient time for skill development and knowledge application (1.1)
AI-generated materials that don't reflect training product requirements or current industry practice (1.1)
Students using AI in assessments without demonstrating their own skills and knowledge (1.3)
Assessment evidence relying on recorded video where live observation is required (1.4)
Assessment validation failing to consider AI-generated assessment materials and outputs (1.5)
AI-generated RPL evidence (1.6)
AI-generated responses to the Annual Declaration on Compliance that don't accurately reflect actual RTO practices (Compliance Requirement 15)
AI-generated or modified qualification templates and documents containing the NRT logo without appropriate review (Compliance Requirements 11 and 13)
There is a common thread: the use of technology must not compromise the integrity, validity or credibility of nationally recognised training.
This is broader than AI. The changes also reinforce the importance of ensuring that delivery arrangements, assessment methods and the evidence used to make decisions genuinely support the requirements of the training product.
Although it repeatedly gets a mention in the overarching themes to the changes, it isn't all AI. So then, within the AI-specific changes, what themes are emerging?
What about AI?
The AI message in the updated Practice Guides is simple:
AI is allowed, but it is not autonomous accountability.
There are six recurring themes across the Practice Guides and Compliance Requirements with regards to use of AI
1. Human oversight and accountability
This is the core theme. AI must not replace the human judgement and accountability needed to maintain quality training, assessment and student outcomes.
Examples:
Human oversight of AI-assisted assessment marking (1.4)
Human involvement in assessment validation decisions (1.5)
Verification of AI-assisted RPL documentation (1.7)
Human review of automated moderation and culturally sensitive matters (2.5)
Human oversight of AI-assisted credential verification (3.2)
Governing persons retaining responsibility for compliance monitoring (4.2)
2. AI governance, risk management and organisational responsibility
RTOs need to understand where AI is being used, who is responsible for it and what risks its use introduces.
Examples:
Documented rationale and risk assessment for AI adoption (4.1)
Clear responsibilities for approving, using and monitoring AI (4.2)
Risk assessment of AI across training, assessment and student support (4.3)
Appropriate controls and data governance for AI-enabled systems
Monitoring third-party use of AI and compliance with contractual obligations (4.2)
3. Integrity and quality of training and assessment
AI must not compromise the integrity of training products, the development of skills and knowledge, or the validity of assessment decisions.
Examples:
Ensuring AI-generated materials reflect training product requirements and current industry practice (1.1)
Ensuring AI does not dilute skill development and knowledge application (1.1, 1.5)
Ensuring students demonstrate their own skills and knowledge when using AI in assessments (1.3)
Including AI-related assessment practices in validation samples (1.5)
Verifying the authenticity of AI-assisted RPL evidence (1.6)
Maintaining appropriate assessment evidence and observation practices (1.4)
4. Student transparency, fairness and protection
RTOs need to consider how AI affects students, their information, their access to services and their ability to participate fairly.
Examples:
Clearly communicating whether and how students may use AI in assessment (1.5)
Protecting student data and privacy when using AI tools (1.7, 1.8, 2.6, 2.7, 2.8)
Identifying bias in AI-enabled screening tools (2.2)
Ensuring AI-enabled platforms accommodate assistive technologies and reasonable adjustments (2.4)
Ensuring students understand when AI-enabled support is being used and how to access human assistance (2.3, 2.4)
Preventing AI-generated content from contributing to cultural harm or discrimination (2.5)
5. Workforce capability and responsible AI use
The introduction of AI creates expectations around workforce capability, industry currency and staff understanding.
Examples:
Keeping pace with industry adoption of AI (1.2)
Providing professional development to maintain trainer and assessor industry currency (1.2)
Ensuring AI adoption does not diminish workforce capability (3.1)
Ensuring staff understand organisational AI policies and can recognise inaccurate, incomplete or unsuitable AI outputs (3.3)
Establishing appropriate visibility of staff AI use where it affects training, assessment and feedback (3.3)
6. Monitoring, evaluation and continuous improvement
AI use should not be treated as a one-off implementation decision. RTOs need to monitor whether it remains appropriate and effective.
Examples:
Evaluating the effectiveness of AI-supported delivery (1.2)
Reviewing AI-related risks through assessment validation (1.5)
Monitoring the quality, effectiveness and appropriateness of AI use (4.4)
Maintaining evaluation reports, monitoring plans, de-identified datasets, action registers and change logs (4.4)
Reviewing emerging risks relating to bias, privacy and transparency (4.2)
The V2 Guidance does not establish a blanket prohibition on AI, nor does it say that every AI application requires the same controls. The expectations are contextual: the level of oversight, risk management and assurance needs to reflect what the AI is being used for and its potential impact.
Want to know more? See the changes in the Training Tools Information Guide: Known Risks – What Changed in ASQA’s V2 Practice Guides. The document gives you a quick, easy way to pinpoint what the changes from V1 to V2 are.
Get it from the Training Tools store.
This article is AITAS Rating 3 - AI-assisted thematic analysis; Copilot-generated images
What is the AITAS?
6.11.26
