What has been reported, and by whom
Featured
‘Disaster waiting to happen’: Parents demand schools revise social media policy
An SMH exclusive reporting that parents are demanding the NSW Department of Education overhaul its social media policy after research showed how easily student photos published by schools can be mass-harvested using AI. The department responded that social media is an “important tool” for schools and that it advises a “risk-based approach”, and said there had been no directive for schools to increase posting.
“Do they not realise that all the photos they publish publicly are so easily data harvested by anyone? It’s a disaster waiting to happen.”
Kara Jade, Newcastle parent, quoted in the article
The audit this reporting draws on is published in full, with its methodology and its limitations, on The Audit. The two government documents in conflict are quoted side by side on Conflicting Advice.
Coverage of this research
The featured piece above is the only coverage of this research so far. Further items will be listed here as they run.
The wider story in Australian media
A record of what has been published, not an endorsement of it.
Experts warn that covering a child's face or school logo with an emoji gives false confidence, because AI can reconstruct or face-swap the image and reverse image search can find uncovered photos posted elsewhere, including by schools. Reports that South Australia's public schools chief has reminded principals to review how images of students are collected, approved, published and stored.
Trade coverage of the eSafety advisory framed around school leaders reconsidering which images they publish. Sets out eSafety's recommended response within 24 hours of a school being targeted: collect evidence, report to platforms, communicate carefully with families, and monitor for content resurfacing.
Regional NSW coverage of the eSafety advisory, localised to Port Stephens schools. The masthead contacted several local public and independent schools; the public schools referred enquiries to the NSW Department of Education. Port Stephens MP Kate Washington is quoted.
Trade write-up in Education Today's Technology section reproducing the eSafety release closely, under the standfirst that deepfakes and staff ratings are attacking schools.
The News Corp network's flagship piece on the eSafety advisory, reporting that schools nationally have begun removing identifiable student photos from websites after expert advice that images are being harvested for bullying, blackmail and child sexual abuse material. Reports that two-part consent forms separating internal from external use are likely.
Sunrise segment on the sector response, leading on Goodstart Early Learning banning class photography.
Sets out the advisory's operational recommendations and the factors eSafety asks schools to weigh before posting.
Political and sector follow-up to the advisory. Goodstart Early Learning banned commercial class photography citing manipulation risk, and Employment Minister Amanda Rishworth and Nationals Senate Leader Bridget McKenzie were interviewed on where responsibility sits.
Goodstart Early Learning, with more than 660 centres nationally, has banned commercial class photography citing the risk of photographs beyond its control being manipulated. Quotes UNSW professor Michael Salter on image scraping at scale.
Published the same week as the eSafety Commissioner's advisory to schools, this piece sets out four questions for anyone posting images of children online. It cites the Commissioner's advice directly and the reported rise in imagery harvested from school social media accounts and websites being used to generate AI abuse material.
Nine's digital write-up of the eSafety advisory, built around the Commissioner's quotes and listing the regulator's pre-posting checklist for schools.
An independent Western Australian community publication's write-up of the advisory for a seniors readership.
Education trade coverage of the eSafety advisory. Reports that images harvested from school social media accounts and websites are being turned into deepfakes, face swaps, memes and fabricated stories about staff, and that eSafety is urging schools to weigh privacy risks before posting and to establish consent and complaints processes.
Trade report on the advisory. Member-gated after the opening paragraph.
A Radio National Hour segment on the eSafety Commissioner's July 2026 advisory to schools, covering how photographs posted to school newsletters, websites, Facebook and Instagram can be manipulated into sexualised deepfakes. Features UNSW criminology professor Michael Salter, who directs the Australasian hub of Childlight.
9News television package on the advisory.
7NEWS television package on the advisory, reporting a sharp rise in image-based abuse of school staff and students.
7NEWS's main online article on the advisory, with original reporting from cyber safety educator Paul Litherland and examples of schools hiding faces, using AI-generated fictional pupils in advertising, and password-protecting newsletters.
The most critical broadcast coverage located. Leads on Australia's secondary school principals saying they are terrified and appalled by the advisory, and frames it as a warning to parents as well as to schools.
Reports the eSafety Commissioner's advisory to schools and the volume of complaints behind it, with experts warning it is no longer safe to publish images of students online. Notes that the Herald has observed a sharp increase in schools sharing images of students and teachers to boost enrolments and build community.
Documents a sharp increase in NSW schools, especially public schools, creating Instagram and TikTok accounts featuring students and staff in viral trends, driven by competition for enrolments. Reports schools hiring social media managers and the NSW Department of Education secretary taking part in trending videos, alongside a growing child-privacy backlash.
Coverage of the NSW Auditor General's finding of significant gaps in how the NSW Department of Education oversees student information security and privacy, with principals carrying technical and legal responsibility without sufficient departmental support. Consent to publish images of students is named as one of the four main areas on which schools sought legal advice.
Reports expert warnings that photographs of children posted by Queensland public schools to their social media pages could be harvested to make AI-generated child sexual abuse material. Quotes Childlight's Professor Jon Rouse arguing that current eSafety guidance and Queensland curriculum frameworks on school social media predate the threat.
Gippsland's Rosedale Primary School, believed to be one of the first Victorian state schools to do so, told families it will no longer post recognisable student photographs on Facebook and will move images to its Compass school management system. Includes the Victorian Department of Education policy position and new national guidelines from University of Wollongong researchers.
Cybersafety expert Susan McLean warns that lax photo-permission practices are letting external presenters unlawfully publish images of students to promote their businesses, and are exposing schools to legal risk including family court subpoenas where orders protect particular children. A direct critique of permission to publish consent design: McLean argues informed consent must specify each use rather than relying on bundled or yes-or-no forms.
Brigidine College in Brisbane stopped publishing recognisable photographs of students on its official social media accounts, limiting shareable imagery to side profiles, large groups and rear views, to reduce the risk of AI image manipulation. Griffith University criminologist Danielle Harris backs the change and argues other schools should remove identifiable student photographs altogether.
A first-person column by a communications specialist whose child's school had the photographer take individual student photos and stitch them into a class photo using AI, raising questions about where those images are stored and which AI tool they were fed into.
Reports the first wave of Australian schools removing identifiable student photos from Instagram and Facebook over deepfake fears, naming St Hilda's Anglican School for Girls in Perth and St Joseph's College in Echuca. Includes a Mentone Grammar principal predicting schools will soon use AI-generated images in place of real students for marketing.
The AFP and child safety experts warn that back-to-school photographs containing school logos, street signs and student names can be used to build a profile of a child for grooming or to create sexually explicit deepfake material. Includes Alannah and Madeline Foundation survey data and comment from the Australian Primary Principals Association.
Mainly about parental sharing after the under-16 ban, but contains a substantive section on schools continuing to publish children's images. Quotes Kirra Pendergast on the contradiction of schools posting student photos to platforms students are banned from, and Tiffani Apps on research into how schools share children's data on public social media pages.
A feature built around Melbourne lawyer and AI ethics expert Kobi Leins, whose daughter was removed from a SEEK holiday tech program after Leins struck out the photo and video release clause on the consent form. The piece argues Australian law leaves children's image privacy almost entirely unprotected, drawing on the LAION dataset finding, including images uploaded by schools, and the Victorian deepfake case.
Reports that Meta has admitted using public Facebook data, including photographs of children posted on public pages, to train its Meta AI app. Focused on parent-posted images rather than school-posted images, but establishes the mechanism that applies to any public page.
The eSafety Commissioner calls on schools to report criminal deepfake incidents to police and releases a guide for schools on managing deepfake incidents. Notes that nude images of female students are being created from school formal and yearbook photographs.
Written for parents receiving a school permission to publish form. Explains that Australian privacy regulation requires informed consent before a school publishes a student's personal information, including photographs and video, on any online platform, and sets out three questions parents should ask before signing.
A Wollongong mother ticked the do-not-publish box when enrolling her two children, then found both children's faces in a video on the school's Facebook page because their teachers were unaware of the no-permission status. The school initially asked her to let it go. The piece draws on University of Wollongong researchers Karley Beckman and Tiffani Apps, and quotes an anonymous teacher on the workload of photographing activities and then editing out non-consenting students.
The Australian Federal Police and the eSafety Commissioner advise caregivers to consider what back-to-school photographs they post and who can see them, warning that publicly visible images and the details they reveal, including school name, uniform and location, can be misused. Addressed to parents rather than to schools.
An explainer on the risks and benefits of posting children's photographs online, drawing on Australian survey data. Its school-relevant advice is that parents should review the policies of institutions their child belongs to and check whether those institutions seek children's consent before photographing them.
Examines how Australian schools use children's images on social media and school websites, and what the Australian Privacy Principles require of them. Identifies three gaps in school consent practice: no platform-specific consent options, unclear procedures for families who decline, and minimal involvement of the students themselves.
Follow-up to the LAION-5B finding, examining who bears responsibility when children's images end up in AI training data. Confirms images were taken from places people assumed were private, including school websites and personal blogs. LAION's position is that responsibility sits with whoever published to the public internet.
Human Rights Watch found 190 photographs of Australian children in the LAION-5B dataset used to train AI image generators including Stable Diffusion and Midjourney. Researcher Hye Jung Han told Guardian Australia the images were taken from photo and video sharing sites as well as school websites, including pages schools had assumed were effectively private because they were unlinked and not indexed by Google.
Legal analysis of the LAION-5B finding focused on what Australian privacy law can and cannot do about children's images being taken into AI training data. References schools generally as institutions that photograph children rather than as a confirmed dataset source.
An SBS News In Depth episode on the Human Rights Watch finding. The episode is explicit that a significant share of the images came from school websites, from sections schools had used to share event photographs with parents and believed were not publicly accessible, and that some images carried full names, ages and school information.
AAP wire coverage of the Human Rights Watch finding that photographs of Australian children were taken into the LAION-5B dataset. Notes that some images came from restricted sources including unlisted YouTube videos and school websites, used without parental consent, and that removing images from a dataset does not undo their use in already-trained models.
A Sydney Breakfast interview on the discovery of roughly 190 images of Australian children in an AI training dataset and what AI systems may do with that data. The segment does not name school websites as a source.
An AM segment on the Human Rights Watch finding that Australian children's photographs, names, locations and ages were taken into the LAION-5B dataset without consent. Hye Jung Han describes finding a newborn and girls in swimsuits at a school sports carnival, and notes that once models have trained on the images the harm cannot be reversed.
Reports the Human Rights Watch finding of 190 identifiable Australian children in the LAION-5B dataset used to train Stable Diffusion and Midjourney. The images carried identifying detail including names, locations and ages, and were sourced from YouTube, Flickr, blogging platforms and school websites. Removing links from the dataset does not remove what trained models have already learned.
An investigation into nudify apps and their use against Australian schoolchildren. Central to the school-imagery question is the eSafety Commissioner's explanation that offenders no longer need many source images and can simply harvest them from social media.
An explainer on photo etiquette for children's images quoting ARC Centre of Excellence for the Digital Child researchers, including a section on photographs taken at schools and sports clubs and how parents can request removal or blurring of uniforms and identifying features.
Reports a large-scale study of school Facebook pages which found schools routinely publish photographs of identifiable students alongside their full names, visible to anyone including data miners and facial recognition firms. The underlying dataset is United States schools, not Australian, though the piece ran in The Conversation's Australian edition. It remains the landmark quantification of this practice.
Early trade coverage of school image-sharing and consent practice, reporting that schools were receiving increasing parent complaints about photo sharing and consent procedures, and advising schools to centralise communications through their own channels rather than relying primarily on third-party social media.
Landmark early coverage of NSW Department of Education permission to publish practice. Forms for major departmental events were changed from a separate opt-in for name and photo publication to a blanket acknowledgement that participation includes publication to social media. Wollongong parents who objected were told there was no other option than a blanket rule. A privacy expert warns the approach risks scrutiny from the state privacy body.
Argues principals underestimate the privacy implications of publishing student images. Colin Anson describes founding pixevety after his eight-year-old daughter's photographs were published by her school without parental consent. Frames posting student images on social media as a potential breach risk under the Australian Privacy Principles.
Pre-AI coverage of the image-harvesting risk. A website solicited and published photographs of girls from more than 70 Australian schools, obtained from social media sites and across the internet, without the girls' knowledge. Queensland Police said the site did not appear to contain child exploitation material. NSW Police contacted the state Education Department.
Every item was opened and read, and its date and byline checked, before it was listed. Nothing is here on the strength of a search result. Where a byline or date could not be confirmed, the entry says so instead of guessing.
Working on this story?
The press kit has the statistics, the methodology, both source documents and the dataset on request. Corrections and missing coverage to hello@algorithms.technology.