Child safety, built to the standards it claims

Why we built this

There are roughly 1.5 billion cameras watching the world, and about 90% of what they record is never watched in real time. A school that already owns thirty of them does not have a surveillance problem. It has an attention problem: by the time someone opens the footage, the incident is hours old and the only thing left to do is write it up. That gap is why AMAN exists. The industry's answer to it has mostly been to record more, for longer, at higher resolution. Ours was to make the cameras a school already owns report what is happening while it is still happening — eight seconds from detection to a verified alert in a safeguarding lead's hand, against the four to eleven minutes manual monitoring takes. The harder decision was what the system would refuse to do. AMAN detects safety events: a weapon, a fight, a fall, an unattended bag. It does not recognise emotions, build psychological profiles of students, or score children. We drew that line before the EU AI Act made a version of it law, and we hold it in markets where nobody would require it — because the first question every safeguarding lead and every ministry asks is not how accurate the system is. It is whether this is surveillance of children. A product that has to argue its way out of that question each time is the wrong product. We would rather the answer be structural.