The Truth for RTO Owners
An ASQA-aligned AI use policy is a documented governance framework that defines how artificial intelligence may be used in training, assessment, administration and learner support while preserving human accountability, privacy, fairness, assessment integrity and compliance with the Standards for RTOs.
It is not a blanket prohibition on AI, but a structure for responsible adoption.
That distinction matters. Artificial intelligence is already changing how RTOs develop learning resources, support learners, analyse information and prepare assessment materials. The question is no longer whether your organisation will encounter AI. The question is whether its use will be deliberate, transparent and defensible.
The regulatory environment has shifted accordingly. The 2025 Standards for RTOs, which commenced on 1 July 2025, do not contain a standalone “AI compliance clause”. However, existing obligations relating to governance, training and assessment, learner outcomes, privacy, staff capability and self-assurance still apply whenever AI is introduced.
ASQA’s Principles for the Responsible Use of Artificial Intelligence in VET and its responsible AI guidance provide a practical interpretive framework.
The pivot is precise.
AI should accelerate good educational design, not conceal weak governance.
What an ASQA-Ready AI Policy Must Achieve
A compliant AI policy must do more than list approved software. It must show that your RTO understands where AI is being used, what risks it creates, who remains accountable, and how the organisation will verify that educational and assessment outcomes remain sound.
A strong policy should answer five questions:
- Where is AI being used?
- What is AI permitted to do?
- What is AI prohibited from doing?
- Who reviews and approves AI-supported work?
- What evidence will demonstrate that the controls are operating?
This is the architecture of defensible compliance. The policy is the visible structure, but the real strength sits in the procedures, records, review points and staff capability beneath it.
ASQA’s guidance makes clear that AI may support work, but it must not replace appropriately qualified human judgement in decisions affecting learners. AI can assist with drafting, classification, analysis and feedback. It cannot become the final decision-maker for competency.
That is not a technical distinction. It is the foundation of assessment integrity.
1. Begin With an AI Use Register
The first practical step is to create an inventory of every current and proposed AI use across the RTO.
This may include:
- Drafting learning resources and assessment questions
- Mapping assessment tasks to units of competency
- Generating examples, scenarios or formative activities
- Providing learner-facing support through chatbots
- Summarising learner feedback or organisational data
- Supporting preliminary marking or feedback
- Assisting with validation documentation
- Developing course concepts and curriculum structures
- Producing marketing, administrative or operational content
The register should identify the tool, its purpose, the data entered, the risk level, the responsible staff member and the required approval process.
This is particularly important for RTOs engaged in accredited course development. If AI contributes to a training product, assessment strategy, learning resource or mapping document, the organisation must be able to demonstrate that a suitably qualified person reviewed the output against the relevant training product and regulatory requirements.
AI-generated content is not automatically accurate, current or compliant.
It is simply unverified content until a competent human has examined it.
2. Separate Assistance From Decision-Making
The most important boundary in an RTO AI policy is the distinction between AI assistance and human decision-making.
AI may assist a qualified trainer, assessor or instructional designer by:
- Suggesting assessment scenarios
- Identifying possible gaps in mapping
- Organising evidence
- Drafting feedback for review
- Highlighting inconsistencies in documentation
- Supporting formative learning activities
- Generating alternative explanations for complex concepts
However, the policy should expressly prohibit AI from:
- Making final competency decisions
- Determining whether a learner has satisfied a unit of competency
- Replacing a qualified assessor
- Conducting validation judgements without qualified human involvement
- Signing off assessment outcomes
- Producing final assessment tools without review
- Completing learner work in a way that compromises authenticity
This is where many policies become vague. They refer to “human oversight” without defining what oversight means.
A defensible policy should specify:
- Who performs the review
- What the reviewer must check
- What records must be retained
- When escalation is required
- Who authorises final use
- How errors or unsuitable outputs are corrected
Human oversight is not a decorative phrase. It is a documented control.
3. Build AI Literacy Across the Workforce
An AI policy cannot operate effectively if staff do not understand the systems they are expected to govern.
AI literacy is the practical capability to use, evaluate and govern artificial intelligence within a defined professional context. It includes understanding how AI systems generate outputs, where those outputs may be unreliable, how bias can appear, what information should not be entered, and when professional judgement must override automation.
For trainers and assessors, AI literacy should include:
- Understanding generative AI limitations and hallucinations
- Recognising generic or synthetic learner responses
- Evaluating whether evidence remains authentic
- Knowing when AI assistance is permissible
- Applying privacy and confidentiality controls
- Verifying AI-generated technical or regulatory claims
- Maintaining the integrity of observation and performance evidence
For instructional designers and compliance staff, it should also include:
- Reviewing AI-generated mapping
- Checking alignment with performance criteria
- Testing assessment validity and sufficiency
- Identifying cognitive and accessibility risks
- Documenting prompts, inputs, outputs and approvals
- Maintaining version control for AI-assisted resources
This is not about making every employee a machine learning specialist. It is about ensuring that the people responsible for learning and assessment can exercise informed judgement.
The Assessment Integrity by Design framework is useful here: assessment must reveal the learner’s reasoning, application and performance, not merely the production of polished text.
4. Protect Privacy, Security and Learner Equity
An AI policy must address the data flowing into and out of each system.
Before adopting an external AI tool, an RTO should establish:
- What information the platform collects
- Whether submitted data is retained or used for training
- Where information is stored
- Who can access prompts and outputs
- Whether learner information is being disclosed
- How confidential organisational information is protected
- What happens when a tool changes its terms or functionality
Learner information should never be entered into an AI system without an appropriate privacy and security assessment.
The policy must also account for equity and accessibility. AI adoption can unintentionally disadvantage learners who have limited digital access, lower digital confidence, disability-related access requirements or language barriers.
It is not innovation if the system creates a new barrier to participation.
A sound policy should require reasonable alternatives, human support and clear communication about when AI is being used in the learner experience. Learners should know how to request human assistance or challenge an AI-supported outcome.
5. Connect the Policy to Assessment Review and Validation
The policy should not sit in isolation from your assessment system.
Under the 2025 Standards, assessment tools must be fit for purpose and aligned with the relevant training product. Assessment practices and judgements must also be quality assured through validation by appropriately skilled and credentialed people. ASQA’s Assessment Practice Guide for Quality Area 1 provides the relevant framework.
Where AI has contributed to assessment design, marking support or validation preparation, your records should show:
- What AI was used for
- Which materials it influenced
- Who reviewed the output
- What changes were made
- How alignment was confirmed
- How the tool was tested against the Principles of Assessment and Rules of Evidence
- How final judgements remained with qualified people
The Assessment Integrity by Design article explores why triangulated evidence, direct observation and learner explanation are becoming increasingly important.
AI detection software is not a substitute for sound assessment design. It is not certainty, but probability. A stronger response is to design assessment tasks that require contextual application, oral explanation, practical demonstration and visible reasoning.
6. Make the Policy a Living Governance Instrument
A policy written once and filed away will not provide meaningful assurance.
Your AI policy should include a review cycle linked to:
- Changes in AI tools or functionality
- New regulatory guidance
- Staff and learner feedback
- Privacy or security incidents
- Assessment validation findings
- Complaints and appeals
- Changes to training products
- Emerging risks identified through self-assurance
The policy should also nominate an accountable owner. This may be a compliance manager, executive leader, academic director or designated AI governance lead, depending on the size of the RTO.
The important point is clarity. Someone must be responsible for maintaining the framework, monitoring implementation and reporting risks.
This is the movement from policy ownership to agent governance. As I explain in From “Author” to “Agent Governor”, the future instructional designer is not merely producing content. The role is increasingly concerned with governing systems, setting parameters and protecting the integrity of outputs.
The Policy Is Only as Strong as Its Design
An AI policy should not be written as a defensive reaction to technology. It should be designed as part of the RTO’s broader quality architecture.
That requires alignment between governance, assessment, staff development, learner communication, data protection and continuous improvement. It also requires an understanding of instructional design. As outlined in What Does an Instructional Designer Actually Do?, the instructional designer’s role is not simply to write content, but to build the structure through which learning becomes measurable and meaningful.
This is where I contribute as an ASQA accreditation consultant and instructional design practitioner. I connect course concept development, accredited course development, assessment strategy, mapping, learning resource design, AI literacy and regulator-facing evidence into one coherent framework.
The goal is not to slow innovation.
The goal is to make innovation trustworthy.
A Practical Final Checklist
Before approving your AI use policy, confirm that it:
- Defines AI and its scope of application
- Includes an organisation-wide AI use register
- Separates permitted assistance from prohibited decision-making
- Requires qualified human review and sign-off
- Protects learner privacy and confidential information
- Addresses bias, accessibility and digital equity
- Defines acceptable learner use of AI
- Connects AI use to assessment review and validation
- Requires evidence of decisions, approvals and changes
- Includes staff AI literacy requirements
- Assigns clear governance responsibility
- Incorporates monitoring, incident response and periodic review
The truth for RTO owners is straightforward: ASQA is unlikely to be satisfied by a policy that merely announces support for ethical AI. The regulator will be concerned with how the policy operates in practice and whether your organisation can demonstrate that operation through reliable evidence.
It is not innovation versus compliance.
It is disciplined design versus unmanaged risk.
If your RTO is developing an AI policy, revising assessment systems or preparing an accredited course for regulatory submission, I can help you build the underlying architecture: clear, rigorous and fit for purpose. My work covers instructional design, accredited course development, assessment mapping, learning resource design, assessor tools, AI literacy and ASQA-facing compliance documentation.
The future belongs to RTOs that can adopt emerging technology without surrendering human judgement.