AI Risk Classification for Australian Universities: OAIC and EU Act

Australian universities deploying AI must navigate domestic privacy law, TEQSA expectations, and EU extraterritorial reach

Australia does not have a dedicated AI regulation equivalent to the EU AI Act. What it has is a Privacy Act 1988 undergoing significant reform, the Australian Privacy Principles (APPs), active guidance from the Office of the Australian Information Commissioner (OAIC), and a Tertiary Education Quality and Standards Agency (TEQSA) that is paying closer attention to AI governance in registered higher education providers.

For Group of Eight universities and other institutions with large EU student cohorts — common across Australian research universities — the EU AI Act's extraterritorial provisions add a third layer of obligation that many governance teams have not yet mapped.

This guide cuts through the complexity and gives Australian higher education institutions a working risk classification framework and a practical compliance roadmap.

The Australian AI regulatory landscape

Privacy Act 1988 and the Australian Privacy Principles

The Privacy Act 1988 and its 13 Australian Privacy Principles govern how APP entities — including universities — collect, use, disclose, and secure personal information. Four APPs are directly relevant to AI deployment:

• APP 1 (Open and transparent management): Universities must have a clearly expressed and up-to-date privacy policy that explains their use of personal information in automated systems. A privacy policy that omits AI processing of applicant data does not meet APP 1.
• APP 3 (Collection of solicited personal information): Collection must be reasonably necessary for the university's functions. AI systems that aggregate personal data beyond what is necessary for the stated admissions or teaching purpose breach APP 3.
• APP 5 (Notification of collection): At or before the time of collection, universities must notify individuals of the purposes for which their information will be used — including AI processing.
• APP 11 (Security of personal information): Universities must take reasonable steps to protect personal information held in AI systems from misuse, interference, loss, and unauthorised access.

The Privacy Act reform process — which will introduce a direct right of action for individuals and strengthen enforcement — is progressing. Universities should build AI governance systems that meet the standards the reformed Act will require, not just the current baseline.

OAIC guidance on AI and privacy

The OAIC has published guidance on AI and privacy that takes a risk-based approach closely aligned with international standards. Key OAIC positions:

• Privacy impact assessments (PIAs) are expected before deploying AI systems that process personal information at scale or in ways likely to have a significant impact on individuals. This is not yet a legislative requirement under the current Privacy Act, but OAIC has signalled it will become mandatory under the reformed Act.
• Automated decision-making transparency: Individuals should be informed when AI substantially influences decisions about them, and should be able to seek human review.
• Algorithmic bias: The OAIC expects organisations to assess and mitigate bias risks in AI systems that make consequential decisions — exactly the kind of decisions admissions, financial aid, and academic progression systems make.

TEQSA and AI governance

TEQSA registers and regulates higher education providers against the Higher Education Standards Framework. While TEQSA does not prescribe AI-specific technical standards, its Governance Standards and Academic Integrity Standards have direct AI governance implications:

• Governance Standard 6 requires governing bodies to oversee strategies for managing institutional risk — which now includes AI risk.
• Academic Integrity provisions require institutions to maintain the integrity of academic assessment, which is directly implicated by AI proctoring and AI plagiarism detection tools.
• TEQSA's regulatory approach is risk-proportionate: institutions that cannot demonstrate coherent AI governance face heightened scrutiny in renewal processes and investigations.

Go8 universities and other research-intensive institutions with complex AI deployments should treat TEQSA renewal cycles as a forcing function for completing AI risk inventories and governance frameworks.

EU AI Act extraterritorial exposure for Australian universities

The EU AI Act (Regulation 2024/1689) applies extraterritorially when AI outputs affect individuals in the EU or when systems are aimed at the EU market. Australian universities — particularly Go8 members with significant European student cohorts and active recruitment in EU countries — face direct exposure.

Annex III, point 3(a) classifies as high-risk any AI system used to "determine access to or admission to educational and vocational training institutions." An Australian university using an AI admissions screening tool that evaluates applications from German, French, Dutch, or other EU-domiciled students is deploying a high-risk AI system under EU law.

Article 29 deployer obligations include:

• Operating the system in accordance with provider instructions
• Implementing human oversight capable of overriding and correcting AI outputs
• Logging outputs and retaining records
• Informing applicants that a high-risk AI system contributes to admission decisions
• Completing a fundamental rights impact assessment before deployment

Universities with active VTAC, UAC, or direct-entry processes that receive EU applications should map EU enrollment volumes and assess whether their AI admissions tools trigger Annex III.

Risk classification for Australian higher education: a working framework

Chatbots, transparency, and automated FAQ handling

Internal Skolbot data shows that 72% of student prospect questions are answerable by automated FAQ — only 7% require human intervention (Source: automated classification of 12,000 Skolbot conversations, 2025). Admissions chatbots that handle the bulk of prospect queries sit in the low-risk category under both Australian privacy law and the EU AI Act's limited-risk tier. The compliance requirement is proportionate: transparency about AI interaction, not a full risk management programme.

Under the APPs and OAIC guidance, this means:

• The privacy policy must disclose chatbot use and what data is collected
• Users must be notified at the point of interaction that they are engaging with an AI system, not a human
• Data collected through chatbot interactions must be used only for the stated purpose and protected appropriately

The practical implementation is simple: "I am an AI assistant for [University Name]. A member of the student enquiries team is available on request." This disclosure fulfils both the OAIC transparency expectation and the EU AI Act Article 50 obligation for institutions with EU nexus.

For the full breakdown of chatbot privacy obligations, see AI Chatbot GDPR Data Collection in Schools. For AI bias risks specific to recruitment, see AI Bias in Student Recruitment.

Practical action plan for Australian universities

1. Build an AI inventory. Document every AI system across admissions, financial aid, HECS-HELP processing, academic integrity, student wellbeing, and marketing. Include vendor tools embedded in your student management system.
2. Classify each system using the APP/OAIC framework and the TEQSA governance lens. Flag any system that makes or substantially influences a consequential decision about an individual as high-risk.
3. Assess EU student exposure. Review enrollment data for EU-domiciled students. If significant, map which AI systems process their applications and assess EU AI Act Annex III applicability.
4. Complete privacy impact assessments for all high-risk AI systems. The OAIC strongly expects PIAs for AI processing at scale; the reformed Privacy Act is likely to make them mandatory.
5. Audit chatbot and automated communication disclosures. Confirm AI disclosure language appears at first interaction on website, UAC/VTAC integrations, and direct communication channels.
6. Review vendor contracts. APPs require universities to take contractual steps to ensure vendors protect personal information appropriately. For EU-exposed systems, require EU AI Act conformity documentation from providers.
7. Document governance decisions. TEQSA assessors and OAIC investigators both look for documented decision-making. An AI risk register with classification rationale, oversight protocols, and PIA references is the right artefact.

For the broader data governance framework, see The EU AI Act and Higher Education and the GDPR Student Data Guide.

Frequently asked questions

Does the EU AI Act apply to Australian universities?

Not as a matter of domestic Australian law. However, it applies extraterritorially when AI outputs affect individuals in the EU or when systems are aimed at the EU market. Australian universities with significant EU student enrollment — particularly Go8 institutions with active European recruitment — face real Annex III exposure. The correct response is to map EU enrollment volumes and assess which AI admissions systems process EU-domiciled applicants.

What does the Privacy Act reform mean for university AI governance?

The Privacy Act reform, which is expected to introduce a direct right of action, stronger enforcement powers, mandatory data breach notification expansions, and mandatory PIAs for high-risk processing, will significantly raise the stakes for AI compliance. Universities should use the current transition period to complete AI inventories, classification frameworks, and PIAs — retrofitting governance after the reformed Act passes is significantly more expensive.

Is biometric data from AI proctoring tools sensitive information under the Privacy Act?

Yes. Biometric information used for identification purposes is sensitive information under the Privacy Act, attracting stronger protection requirements. AI proctoring systems that use facial recognition or keystroke dynamics collect sensitive information and require explicit consent, strict security measures, and clear purpose limitations. TEQSA's academic integrity expectations do not override Privacy Act obligations.

How does HECS-HELP interact with AI data governance?

HECS-HELP processing involves Commonwealth-held financial data about students. AI systems that incorporate or interact with HECS-HELP data must comply with both the Privacy Act and the specific data handling requirements in the Higher Education Support Act 2003. Any AI tool that processes financial aid decisions should be treated as high-risk from a governance perspective.

Our university uses ATAR cutoffs supplemented by AI portfolio assessment. Is that high-risk?

Yes, almost certainly. Any AI system that contributes to decisions about whether an applicant gains admission — even as a supplement to ATAR — is in the high-risk tier under OAIC guidance and EU AI Act Annex III (if EU applicants are involved). The key question is whether the AI output substantially influences the decision. If it does, the full suite of impact assessment, human oversight, and transparency obligations applies.