The AI Surveillance Shift

In October 2022, a student walked into Thomas R. Proctor High School in Utica, New York carrying a large knife. He walked it straight past an AI weapons scanner the district had spent $3.7 million (£2.8 million) on, and later that day he used it on another student. When staff turned the sensitivity up so the system would not miss the next one, it began flagging laptops, binders and water bottles at a rate of roughly one false alarm for every two people scanned. Two years later the Federal Trade Commission concluded that the company had oversold what its technology could actually do. By then the same product was installed in more than 800 schools across 40 states.

That gap, between what one of these systems does in a demo and what it does in your building, is what we want to talk about. AI-driven surveillance is being bought into corporate sites, residences and event venues faster than most teams can write a policy for it. It gets sold as more eyes that never blink. Handled badly, it becomes fewer eyes that nobody trusts, and it takes the blame with it when something goes wrong.

In this article we take autonomous monitoring apart the way you would take apart any new piece of kit before it goes live on your site, and then we hand you the tools to do it yourself. You will get:

It comes with a downloadable scored toolkit, the Autonomous Monitoring Evaluation, so you can put a number on a product and its blind spots before you sign, rather than working it out after an incident.