An academic at the Australian National University has accused the institution of a “hysterical” response to students using AI to cheat, as tertiary institutions across Australia rush to reclassify assessments as “secure” or “insecure” in an attempt to preserve the credibility of their degrees. The Guardian report names ANU, the University of Queensland, and the University of Melbourne as the leading movers of a sector-wide shift toward oral exams, in-person assessment, and explicit AI-use declarations.
🔍 THE BOTTOM LINE
The fight is not about whether AI is used — the 2025 Australian Digital Inclusion Index found 78.9% of secondary and tertiary students already use generative AI. The fight is about how universities respond, and whether the response is measured or panicked. ANU’s own staff are split down the middle: one side calls the consultation “earnest”; the other calls it “a hysterical response” that could roll back inclusion gains for students with disabilities and caring responsibilities.
What is a “secure” assessment?
A secure assessment is any evaluation designed so that a student cannot meaningfully outsource it to an AI. In practice, this means in-person exams, oral vivas, supervised lab work, or — at the technical frontier — assessments delivered on locked-down devices that block AI access during the task. An insecure assessment is the opposite: take-home essays, open-book reports, and anything submitted outside a controlled environment. The binary is the framework ANU, UQ, and Melbourne are now adopting as their default. A third category — “open” assessments, where AI use is permitted but the student must critically appraise the AI’s output — is being trialled at UQ under deputy vice-chancellor Kris Ryan.
What ANU Is Actually Proposing
ANU has released a consultation paper to academic and teaching staff with three options on the table, according to the Guardian:
- Classify every assessment as either “secure” or “insecure” — a binary that forces course designers to decide upfront whether a student can complete the task without AI assistance.
- Require students to declare which stages of an assessment AI was used in — a transparency approach that treats AI as a tool to be disclosed rather than banned.
- Move toward greater on-campus assessment — the most resource-intensive option, requiring scheduling, room capacity, and staffing that several ANU academics say the university does not currently have.
The consultation closed ahead of the second semester, and at least one ANU faculty has already issued draft guidelines pushing staff toward more on-campus assessment — ahead of the institutional decision being finalised. That is the gap that produced the “hysterical” accusation: the consultation is still technically open, but the implementation is already moving.
The Internal Split
Two ANU academics, both speaking anonymously to Guardian Australia, gave opposite reads of the same process. The first framed it as a necessary if under-resourced adjustment:
“Everyone is adjusting in a panicked way, which is aggravated by a lack of resources to support teaching. It makes the shift to secure assessment really complicated.”
The second went further, naming the panic itself as the problem:
“This is not an earnest consultation, this is a reactionary response, perhaps one would argue a hysterical response. We are at a point of panic now. That panic has resulted in proposals that could see inclusion go backwards.”
The inclusion reference is specific. Oral exams, in-person assessments, and on-campus scheduling disproportionately affect students with disabilities, students with caring responsibilities, and students who work — the same groups that benefited from the shift to take-home and flexible assessment during and after the pandemic. A pivot back to in-person testing, done fast, can unwind those gains in a single semester.
The National Capability Argument
ANU law professor Will Bateman, who leads research projects on AI regulation, framed the stakes in sharper terms than the internal-consultation debate. His line is the one that has travelled furthest since the Guardian piece was published:
“If we don’t address the erosion of the norms and rigour in our education created by AI, we are just shipping our national intellectual capability to companies in California and China.”
Bateman’s point is that the credential — the degree itself — only has value if outside employers, regulators, and other universities can trust that it represents genuine student capability. If AI use hollows out that trust, the credential inflates while the underlying capability deflates. The beneficiary of the gap is not the student; it is the AI provider whose tool the student used. “Shipping our national intellectual capability” is a pointed way of saying that the value transfer goes from the Australian education system to the companies that built the models.
Bateman also pushed back on the assumption that “secure” must mean “in-person en masse”:
“Responding to AI doesn’t automatically mean sending students back to exam halls en masse, but we need serious investment in fully-secure devices to ensure that students are actually learning the material we teach and to assure people outside the university that the degrees we issue are meaningful.”
What the Other Universities Are Doing
The Guardian report confirms three parallel responses across Australia’s Group of Eight:
- University of Queensland — began implementing “secure” assessment policies earlier in 2026, but drew criticism for scheduling in-person oral assessments during night hours and over weekends. Deputy vice-chancellor Kris Ryan told the Guardian the university has “three teams working now” on scheduling, assessment types, and student alternatives.
- University of Melbourne — moving toward secure assessment with a specific emphasis on interactive oral exams. Deputy vice-chancellor Gregor Kennedy framed it as “transforming our assessment approach to embed secure assessment types that ensure integrity and verify students’ work amidst the rise of AI tools.”
- Australian Tertiary Education Commission — the new head, Barney Glover, says he is seeing more universities reintroduce oral exams, but warned institutions will “need to be really responsible” in how they use or restrict AI. He expressed confidence the sector will adapt.
The pattern across all three is the same: a drift back toward oral, in-person, and supervised assessment as the only formats that can be reliably protected from AI use. The 2025 Index data — 78.9% of secondary and tertiary students using generative AI — makes the drift almost inevitable. The question is whether it happens as a measured redesign or as the panic one ANU academic described.
NZ Angle
New Zealand is roughly six months behind Australia on this curve, but moving the same direction. The University of Auckland and Victoria University of Wellington have both updated academic-integrity policies to address AI use in 2025–26, with Victoria introducing a secure-assessment framework similar to ANU’s binary model. NZQA has issued updated NCEA internal-assessment guidance that explicitly names generative AI as a risk to internal assessment credibility. The Tertiary Education Commission’s 2025 statement on AI in tertiary education takes a lighter-touch approach than Australia’s, but the direction of travel — toward oral, in-person, and AI-disclosed assessment — is the same. For NZ readers, the ANU fight is a preview of the same debate landing here in the next academic year, with the same inclusion-versus-integrity trade-off.
The Real Question
The “hysterical” framing from the ANU critic is not a defence of AI cheating. It is a warning that a panicked response can cause more damage than the problem it is trying to solve — specifically by rolling back assessment formats that made universities accessible to students who cannot sit a 9am Monday invigilated exam. The sector-level question is whether universities can design secure assessments that preserve both integrity and inclusion, or whether the pressure of the moment forces a binary choice between the two. Bateman’s “shipping intellectual capability” line is one half of the answer. The “hysterical” quote is the other. Both will be quoted in the next round of NZ and Australian vice-chancellors’ meetings.
❓ FAQ
What exactly did the ANU academic call “hysterical”? The consultation paper proposing three assessment classifications — secure, insecure, and AI-declared — and the speed at which faculties were being pushed to implement on-campus assessment before the consultation had concluded. The critic’s concern was that panic was overriding inclusion considerations for disabled students and carers.
How widespread is student AI use in Australian universities? The 2025 Australian Digital Inclusion Index found 78.9% of secondary and tertiary students use generative AI. That figure is the baseline driving the sector-wide pivot to secure assessment — universities are no longer designing policy for the possibility of AI use, but for the reality that nearly every student is using it.
Are oral exams actually secure against AI? Yes, in the immediate sense — a live oral viva cannot be outsourced to ChatGPT in real time. The trade-offs are cost (oral exams require examiner time that scales poorly to large cohorts), scheduling (UQ’s night-and-weekend oral exam slots drew criticism), and accessibility (oral formats can disadvantage some disabled students and non-native English speakers).
Is New Zealand doing the same thing? Not yet at the same scale. Victoria University of Wellington has introduced a secure-assessment framework similar to ANU’s. The University of Auckland and NZQA have updated AI-related guidance. NZ’s Tertiary Education Commission has taken a lighter-touch stance than Australia’s, but the direction of travel is the same.
What does “shipping intellectual capability to California and China” mean? Bateman’s phrase refers to the value transfer that happens when students outsource their cognitive work to AI models built by companies in California (OpenAI, Anthropic, Google, Meta) and China (DeepSeek, Alibaba, Zhipu). The student gets the output; the model provider gets the usage data, the capability signal, and the dependency. The degree-holder’s genuine capability erodes while the AI provider’s position strengthens.
🔍 THE BOTTOM LINE
Australia’s top universities are quietly rebuilding their assessment systems around a single assumption: that take-home essays are no longer a reliable measure of student capability. ANU’s internal fight is the early warning that the rebuild, done under panic and without resourcing, can break the inclusion gains the same institutions spent the pandemic years building. The ANU law professor’s line — “shipping our national intellectual capability to companies in California and China” — is the sharpest framing of the stake. The “hysterical” line is the sharpest framing of the risk of getting the response wrong. Both will travel.
