91av

I AM one of millions of people around the globe who have a smart doorbell with a hidden camera, controlled through an app. Now I can’t stop watching videos of my porch and the street beyond. I have a Google Nest smart home system, and it sends alerts to my phone that say things like: “Your doorbell camera thinks it has spotted a person.” Who wouldn’t be curious after receiving that? Is the person on my porch a suspicious lurker? Someone delivering a package? An old friend? In my case, the answer has, so far, been none of these. Usually, the people my camera has spotted are my neighbours’ children racing out the door to school, or a person walking by with their dog.

It doesn’t feel like surveillance, exactly. There is something almost fun about peering out of your front door without anyone knowing. Yet hundreds of police departments in the US are using footage from Amazon’s Ring doorbells to snoop on streets. They ask residents to install Ring to create a of street cameras that law enforcement can access. and freebies to people buying Ring through the police.

Officers need user consent to access footage through Ring’s social app, Neighbors, but they don’t need a warrant. That means I could spy on my neighbours and passersby for the police without any court oversight. My Google Nest system doesn’t currently have an app like Neighbors, nor partnerships with police, but it is still easy to share footage. Though some cities like San Francisco have banned police from using facial recognition, many places and organisations are embracing it.

There is a big difference between smart doorbells and traditional CCTV. For one thing, they are privately owned. In the US, that means police can get all kinds of surveillance data just by asking nicely through apps like Neighbors.

I am constantly uploading data about people’s faces and whereabouts to the cloud. Each time a person rings my bell, my app takes a snapshot and asks whether I know them. I can choose to label friends’ faces with their names, or simply say that I recognise them so Nest can alert me when a “known person” arrives. My camera also captures images of cars driving on my street and the junction nearby, which means my doorbell could be logging number plates too.

Where is all that data going? We can be sure that at least some of it is used to train algorithms to recognise people. Companies are desperate for this kind of face data. A Google contractor was recently busted for paying homeless people if they agreed to pose for selfies in a “game”. That game turned out to be . And there are other sneaky ways to harvest faces. Security at a Taylor Swift concert in 2018 set up a kiosk that secretly . Several concert venues and airports have similar technologies in place.

“Facial recognition algorithms are notoriously unreliable – especially for people of colour”

What’s more, facial recognition algorithms are notoriously unreliable – especially when it comes to people of colour. Due to biased data sets and other technical problems, they misidentify people of colour far more often than white people. This means police often nab the wrong person. But tech companies are working to close the face divide. One reason why Google hired that contractor to take pictures of the homeless was reportedly to get more people of colour into its database.

Chinese companies who make surveillance software are taking things to the next level. In 2018, China-based company CloudWalk Technology cut a deal with the government of Zimbabwe to do . Recently, Chinese authorities have taken , a primarily Muslim minority group, in the hopes that they will learn how to recreate Uighur facial features from DNA. The Chinese government has put roughly a million Uighurs into . It hopes to simplify the process by using DNA to identify Uighurs in surveillance footage.

I just had a conversation with my partner through the doorbell. He was fixing the lock on our door, and my phone alerted me to a “familiar face”. When I saw him on camera, I turned on its speaker and yelled “EXTERMINATE! EXTERMINATE!” Somehow, he guessed it was me. It’s harmless footage, but who knows what else was captured behind him? I have no idea what will happen to my Dalek impression video if Google sells Nest to a surveillance company, and I can’t control who might be targeted inadvertently by my doorbell’s constant vigilance. Even when I tell my app to delete footage weekly, I’ll never feel completely sure.