Can AI catch a thief? New technology could help report suspicious activity in real time

We already use artificial intelligence, or AI, to make our lives easier through personal assistance like Apple’s Siri and in self-driving vehicles but now hundreds of stores nationwide are using it to spot shoplifters.

“We detect when someone is doing a suspicious gesture, or a suspicious behavior and we send an alert,” says Pablo Blanco from Veesion, a French tech company that connects to the cameras stores already have.

Veesion runs camera images through an algorithm programmed to alert management of suspicious activities in real time.

“When someone takes something from an aisle, they put it inside their pockets, inside the trousers, inside their jacket…when someone is consuming inside the store, someone is drinking. Someone is eating inside the store,” says Blanco.

The algorithm also sends an alert when it detects someone putting an item inside a purse, backpack, or a baby stroller.

The idea for stores like Key Food in Hollywood is that they can free up employees to help customers on the floor.

“Having to sit and watch the cameras, that's not really a feasible task to do all day. So, I think that this helps us eliminate that,” says Key Food Store Manager Nicole Espinal.

To see how quickly the system works, reporter Myriam Masihy posed as a client looking for a deep discount and seconds after she placed two olive jars in her purse, the manager got an alert with video of the pretend theft.

The same thing happened in California when an NBC Bay Area colleague pretended to steal a bottle of wine.

Veesion’s technology is already operating in supermarkets, pharmacies, liquor, and electronic stores.

“We already have 3000 stores around the world. Here in U.S., we have 300 already,” Blanco says.

To comply with privacy and consumer protection laws and prevent the algorithm from discriminating against a race, gender, or protected class, Veesion says it doesn’t store biometric information, only gestures.

“Just to give you an example, someone steels in the store, we send an alert. The next day, the same guy comes back to the store, we don't send an alert. If he steals, we send a new alert, but we don't identify,” Blanco explains.

Veesion is just one of many new anti-theft tools being used by retailers.

At the University of Florida’s Prevention Research Council, there’s a team of scientists working with 110 tech companies and 18 retailers on anti-theft technology.

“We’ve got 230 to 250, depending on who you ask, technologies in here,” says Director Read Hayes, a criminologist showed us the mock store they use as their lab, complete with shelves and such. “There’s A-I in here that can pick up whether you’re not really scanning an item at self-checkout,” says Hayes.

While they put those systems to the test, managers at Key Food in Hollywood say the new system has been eye opening.

“We probably originally thought maybe there was one or two thefts a week, and I'm starting to think it's more than that,” says Espinal.