Greens move on smart glasses as teachers defer to AI marks

Digital Attitudes, 29 August 2026

Saturday of Term 3 Week 7, and this week added three fronts to the Australian AI in schools file that were not on the whiteboard last Monday: what a student can smuggle into an exam room, what a teacher rubber stamps at the marking desk, and what a defendant did with a school website's photo gallery.

Greens move on smart glasses as the next cheating vector for Australian classrooms

The Australian Greens will introduce a bill when Parliament resumes in September to impose a 12 month import ban on camera enabled smart glasses (Meta and Anko among them), pending stronger Privacy Act protections that would make non consensual filming from a wearable recording device a breach and give the Information and eSafety Commissioners investigation and penalty powers (ABC News on the push for an import ban on smart glasses, EducationHQ on the push to halt imports of covert recording tech glasses). A Monash study cited in coverage found significant proportions of owners admitted to covertly recording in changerooms, on beaches and in workplaces. CQUniversity Professor Ken Purnell put the ed tech framing on the record: students "can simply look at a question, have it processed by AI, and receive an answer almost instantly." His summary of the paradox reads, "better cheats, but poorer learners." Senator David Shoebridge cited one third public support for an outright ban, and carve outs would allow use as an accessibility aid.

Why this matters: for a NSW head of teaching and learning already recasting Term 4 assessments under NESA's proposed take home ban (Wednesday's brief) and for a Victorian principal running BYOD in a secondary science lab, smart glasses are a new hardware vector that a device policy written for phones does not cover. Two Term 4 tasks. First, update the exam and assessment device policy to explicitly name wearable recording devices alongside phones and smartwatches, because a policy that only names phones will not survive a formal integrity dispute. Second, cost the invigilation change: a QR scan or a phone drop table does not spot a $99 Anko frame. The Greens bill is unlikely to pass without government support, so the operational assumption for now is that supply keeps coming. The College Board banned smart glasses at SAT centres this month and Polk County (Florida) has banned them on campus, so the sector is already moving without waiting for a Privacy Act amendment.

A Monash led study says teachers defer to AI grades, and the confident ones defer most

Professor Rigissa Megalokonomou (Monash Business School), Dr Sofoklis Goulas (Yale) and Dr Panagiotis Sotirakopoulos (Curtin University) published Flawed Oversight: Teachers Don't Catch AI's Grading Mistakes When AI is Harsh in PNAS Nexus (Volume 5, Issue 6, June 2026), and The Educator K/12 surfaced it in Australian coverage this week (The Educator K/12 on AI generated grades less likely to be challenged by teachers, Mirage News on the Monash finding that people trust AI over humans even when it's wrong). The design is a randomised experiment with more than 1,300 Greek teachers grading identical student work, with the recommended score labelled as AI or human. When the recommended score was deliberately harsh, teachers accepted the AI attribution at a rate 22 per cent higher than the same score attributed to a colleague. The counterintuitive finding: younger teachers, teachers with postgraduate qualifications, and teachers who described themselves as technologically confident were the least likely to push back on a harsh AI grade. Megalokonomou on the record: "the very people we might expect to be the most capable and critical users of AI tools turned out to be the most likely to defer." The Educator's separate framing figure: 48 per cent of Australian teachers now use AI weekly for lesson preparation, and only 16 per cent actively encourage colleagues to adopt it.

Why this matters: this is the specific failure mode the AI marking pitch has been quietly assuming does not happen, and it lands in the same fortnight the Turnitin 53.6 per cent ANZ tertiary submission figure is still being used to justify vendor AI marking rollouts (July's brief). Two operational reads for a head of teaching and learning approving an AI marking or AI feedback pilot this term. First, the sample bias in the buying decision is now documented: the teachers who volunteer for an AI marking pilot are exactly the cohort the study says will defer most, so a pilot with 20 self selected teachers will underweight the disagreement rate that a full staff rollout will surface. Second, the moderation process a school runs alongside an AI marking pilot needs a specific defer test: does the moderator see the AI score before setting their own. If the answer is yes, the study says the moderation is contaminated. Any 2027 vendor conversation about AI marking should be asked to name how their product blinds the moderator, because the alternative is the 22 per cent gap sitting inside the school's reported grades.

The Rotondo trial in Brisbane puts the school photo supply chain in the dock

The Crown opened its case against Antonio Rotondo, 56, in Brisbane District Court on Tuesday 25 August against 17 charges including stalking, distributing intimate images, and distributing child exploitation material, with prosecutor Ben Jackson telling the jury Rotondo used AI tools to strip clothing from public photographs of female students and staff from a Queensland private school and models from a fashion show, then emailed the composites to two Brisbane private schools in September 2023 (ABC News on the Rotondo Brisbane deepfake images trial, EducationHQ on the alleged AI schoolgirl pictures). One image the Crown says Rotondo modified is of a 13 year old girl. The prosecution has put Rotondo's own statements to police on the record: "it's fun to just press a button and remove their clothes. It's like X ray vision glasses" and "every woman on Earth should live in fear of a camera." Rotondo has pleaded not guilty and the trial continues. He was fined $343,500 in the Federal Court in September 2025 in a separate matter for deepfake images of six prominent Australian women, prosecuted by the eSafety Commissioner.

Why this matters: eSafety's late July advisory already told Australian schools to review what they publish, and the Rotondo trial is the current criminal case that puts the operational risk in front of a school council rather than in front of a policy officer. Two Term 4 tasks. First, a school communications audit, not for whether the photos exist (they do, on every school website and enrolment brochure in the country), but for whether the metadata around them makes a supply chain easier for a defendant. Named Year 8 team photo galleries, high resolution staff head shots on public bios, alumni pages with class years, prize night videos linked from the school YouTube channel: each is a design decision an operations manager can revisit under the OAIC Children's Online Privacy Code. Second, the parent communication template for what a school says when the images circulate, because a script written now will read better than one written after a Monday morning email from a parent who received one. The Rotondo trial is running because a school photo library was scraped. Every Australian principal has a school photo library.

Three stories, one Saturday, and the throughline is that Australian AI in schools guardrails are now being drafted at the hardware layer (smart glasses), the marking desk (deferred oversight), and the school website (photo supply chain), and each of them is a policy line an operations manager rather than a strategy officer signs off. Have a slow Saturday. Coverage resumes on Sunday, and prior editions live at https://digitalattitudes.com.au.