Wearable AI is changing the exam-security conversation before schools are ready

Exam security leaders are increasingly describing AI cheating as a hardware problem, not just a software one.

That shift became harder to ignore this week when Ofqual warned students not to bring phones or smart devices into exam halls as the 2026 exam season opened in England. Days earlier, Times Higher Education reported on researchers arguing that AI-enabled glasses, earbuds, rings, and other wearables are eroding assumptions behind traditional invigilation. Put together, those warnings suggest the assessment problem is moving beyond obvious devices like phones and watches.

The concern is not just that students can access more powerful tools. It is that the tools are getting smaller, cheaper, and easier to disguise. Policies written for an earlier generation of devices may explicitly ban phones while saying little about eyewear, discreet audio devices, or accessories that do not immediately read as technology. That widens the gap between what the rules describe and what invigilators can realistically detect in real time.

The pressure is now on assessment design as well as supervision

This is why the conversation is moving beyond proctoring. The easier it becomes for students to receive covert assistance, the more pressure exam boards, schools, and tutoring providers will face to use tasks that require explanation, process evidence, or visible working rather than answer production alone. That does not mean every high-stakes exam can be redesigned quickly, but it does mean the risk model has changed.

In the short term, institutions still need practical fixes. Staff need clearer guidance on prohibited devices. Students need simple, early reminders before exam day. Malpractice policies need to cover the kinds of AI-enabled wearables already available in consumer markets, not just the devices adults are most used to spotting.

The underlying issue is broader: the next stage of AI disruption in assessment may arrive not through the browser window, but through what students can carry into the room.

Detection cannot become an endless hardware race

Invigilators can learn to recognize today's smart glasses or wireless earbuds, but a device-by-device strategy ages quickly. Consumer hardware changes faster than exam regulations, and legitimate accessibility tools can look similar to prohibited assistance. More aggressive searches or surveillance may catch some misuse while increasing privacy concerns and the chance that students with accommodations are treated with suspicion.

That makes proportionality important. Exam bodies need rules based on capabilities, not brand names, along with a clear process for approved devices. Schools need practical training that helps staff identify concerning behavior without asking them to become electronics experts. Students should know before arriving which devices must be removed and how an authorized exception will be verified.

The technology also puts pressure on the evidence used in malpractice cases. A suspicious object is not the same as proof that a student received assistance. Institutions need procedures for securing a device, recording what happened, and allowing an appeal. As tools become less visible, the temptation to rely on inference will grow. A defensible system has to protect exam integrity without weakening procedural fairness.

Assessment design is the longer-term response

No redesign makes cheating impossible, and high-stakes systems cannot replace every exam at once. But tasks that require intermediate reasoning, oral explanation, local application, or authenticated work produced over time reduce the value of a hidden answer feed. That shift has costs: marking can take longer, standardization becomes harder, and schools need staff and infrastructure to deliver more varied assessment.

Tutors and teachers have a role in preparing students for that environment. The useful response is not simply another warning about punishment. Students need practice showing how they reached an answer, checking machine output, and understanding when assistance becomes misrepresentation. Those are learning goals as well as security measures.

Wearable AI does not make conventional exams obsolete overnight. It does remove the comfort of treating the exam hall as a technologically sealed room. The institutions that respond well will combine clearer device rules, fairer evidence procedures, and selective redesign. Those that focus only on spotting the latest gadget will always be preparing for the previous version of the problem.