SUBMIT A SUPPORT REQUEST
A parent in Ray-Bans walks into your reception. Now what?
Two Harvard students built a tool that turns Ray-Ban Meta glasses into a doxxing device. They wore the glasses around Boston, looked at strangers, and within seconds had names, home addresses, phone numbers and in some cases family members and partial social security details. They called it I-XRAY, wired the glasses to the facial recognition service PimEyes, and published the demonstration deliberately to force a public conversation about wearable surveillance.
The same model of glasses is now on the shelves of UK opticians and on every Ray-Ban concession in central London. The conversation the students wanted to force has reached Westminster, the courts and the regulator. It has not yet reached most school safeguarding policies.
What has happened since the demo is harder to dismiss. In March 2026 the Information Commissioner’s Office wrote to Meta after Swedish journalists revealed that contractors at the data annotation firm Sama in Nairobi had been reviewing intimate footage captured through the glasses, including users undressing, using the toilet, handling bank cards and engaging in sexual activity. The ICO described the disclosures as concerning and demanded urgent clarification on how Meta was meeting its UK data protection obligations. A US class action followed in the same week. The potential UK fine, if a GDPR breach is confirmed, sits at up to four per cent of global turnover.
Why this matters to primary schools
A school is one of the few buildings in the country where the consent assumptions baked into the glasses become unworkable. Children cannot meaningfully consent to having their faces processed by AI. Staff cannot consent to being filmed in front of pupils. Visitors do not know what other visitors are wearing. The Department for Education updated its mobile phone guidance in January 2026 and made the expectation clear that schools should be phone-free environments by default from 1 April 2026. That guidance does not name smart glasses. The Ofsted post accompanying it does not name smart glasses. The category schools have spent eighteen months designing policy around is the rectangular device in a child’s blazer pocket, not the recording device on a child or adults nose.
The one published example of a UK school addressing this directly is Shrewsbury School, which on 21 November 2025 banned smart glasses for pupils, staff, parents and visitors. The letter from Assistant Deputy Head Henry Exham named discreet recording, livestreaming and instant sharing as the primary risks and was explicit that the technology potentially breaches UK GDPR and the Data Protection Act 2018. That is a private senior school. Its decision was published eight months before the DfE chose not to mention the technology in updated statutory-aligned guidance. Most maintained primary schools we work with have no equivalent provision.
Smart glasses can capture images or video without consent and may inadvertently reveal sensitive or identifying information about pupils, staff, or visitors.
Shrewsbury School letter to parents, November 2025
What this could mean in practice
Three implications deserve to sit with school leaders for longer than the time it takes to read a blog post. The first is the changing room and toilet question. Phone policies were written for a device a child has to hold up and aim. Smart glasses are aimed by where the wearer is looking, and the recording indicator can be obscured with a fingertip. A swimming gala. A Year 6 residential. A school production with shared dressing rooms. Parents in changing room corridors at sports day. The existing safeguarding language is not designed to refuse the device, because no one anticipated needing to refuse spectacles.
The second is the deepfake question. Recent academic work on consumer-grade synthetic media shows that a working likeness model can now be trained on as few as twenty images using around fifteen minutes of GPU time on a domestic computer. A child filmed by a wearer in a corridor for a few seconds produces well over twenty usable frames. The same research documents an ecosystem of nudify apps marketed at younger users that require no technical skill. Australia’s eSafety Commissioner is already publishing case data on deepfake abuse circulating in schools. The supply chain from corridor footage to fabricated content involving a real child or a real teacher has shortened to the point that the slowest part of it is the wearer’s motivation.
The third is the exam integrity question, which has moved faster than most assessment policies. In UAB Business Enterprise v Oneta Limited [2026] EWHC 543 (Ch) the High Court rejected the evidence of the claimant after the judge found he had been coached through smart glasses while giving testimony. Judge Agnello dismissed his evidence in its entirety and made an indemnity costs order against him. The implication for schools is that the assumption underpinning oral assessments and viva voce style examinations, namely that being seen and heard means being unaided, no longer holds. SATs administration arrangements and GCSE practical assessments were not written with that in mind. Some of the language schools rely on around invigilation predates the existence of a camera that does not have to be held.
What we’re thinking about at itchyrobot
We have spent the last year writing about AI tools in classrooms on the assumption that the AI was the thing a school was choosing to introduce. The smart glasses story breaks that assumption. The technology arrives on a parent’s face at a parents’ evening, on a sixth former visiting younger siblings at pickup, on a contractor doing electrical work in the corridor outside Year 4. There is no procurement decision to make. There is no opt-in to negotiate. The school becomes a data controller in scenarios it never agreed to become a data controller in, because Meta’s voice-data settings are buried inside individual user accounts and the bystanders being recorded were never asked.
The question is whether the next twelve to thirty-six months in schools will be defined less by what schools choose to build with AI, and more by what AI builds about schools without permission. The accessibility argument complicates that further, because the live captioning and translation features of these devices have a real benefit for EAL pupils and pupils with dyslexia, and a blanket ban without a position on that benefit creates an equality duty gap that an inspector will find before a school does.
None of that produces a tidy answer. It produces a question that we think school leaders, designated safeguarding leads and chairs of governors will be answering for the rest of this decade whether they choose to or not. Search your safeguarding policy for the words smart glasses. If they are not there, the gap is not a drafting oversight. It is a position by default.
