Future Tense

Say Hello to Dog Facial Recognition Technology

Should you worry about your pet’s privacy, yours, or both?

A dog's face mapped by facial recognition technology.
Photo illustration by Slate. Photo by Holly Allen.

As facial recognition has taken off, so have privacy concerns. There is no shortage of recent scandals: China’s government is actively monitoring citizens through a network of cameras and algorithms matching images to people, and U.S. Immigration and Customs Enforcement used facial recognition to analyze driver’s license databases without citizens’ knowledge. As a result, legal and privacy scholars, tech columnists, and even government officials have cautioned against the technology. So far, most of the discussion of facial recognition has centered on human faces, but it’s on the rise among other animals as well—and with that comes a different set of privacy concerns.

Perhaps the most popular application of animal facial recognition technology is creating databases of pets. Apps like PiP and Finding Rover use facial recognition technology to match pictures of lost dogs and cats with new shelter arrivals or found pets to reunite them with their owners. Finding Rover’s founder and “chief barkeroo” (or CEO) John Polimeno says that his app is now home to more than 700,000 pet pictures and works with 700 shelters around the world, growing its network by five to seven shelters a week.

In recent local TV segments announcing shelters’ new partnerships with Finding Rover, employees at those partner shelters encourage people to download the app and upload their pets’ pictures. “The more people who use Finding Rover, the more successful it will be,” says the adoption coordinator at the New Albany Floyd County Animal Shelter in Indiana. “The more people involved, the greater impact it will have,” echoes the executive director of the Charlottesville, Virginia, Society for the Prevention of Cruelty to Animals. Getting word out to local audiences seems to be working; Polimeno says that on a typical day, they get 1,000 new registrations. There’s no cost to users or shelters, and once a pet’s photo is uploaded, it stays in the system as a sort of “security blanket” in case your pet does ever end up found by another Finding Rover user or in a partner shelter.

According to Finding Rover, the company’s technology is 98 percent accurate. It’s impossible to independently verify this, but Polimeno’s team worked with researchers at the University of Utah to develop the technology, which uses 138 points on an animal’s face to identify it. The researchers are continuing to tweak the algorithm. When Finding Rover first launched, users had to click on the screen to identify pets’ eyes and nose within the picture they uploaded, but the algorithm is now able to recognize those features and analyze them without user help. PiP also claims 98 percent accuracy for its facial recognition tech, and Chinese A.I. startup Megvii says its new app that identifies pets by their noses is 95 percent accurate.

If those numbers are accurate, there will be a lot of opportunities to reunite pets and owners—but also, potentially, a lot of opportunities to mine user data. I downloaded the app to see how it works, and it immediately asked for permission to use my location “to help us find lost, found, and adoptable pets near you,” which seems reasonable enough. And like most apps, it allowed me to sign up for a new account with a name and email address or log in using a Facebook account. I opted for the former. It autofilled a field with a ZIP code based on my current location, which I manually replaced with the ZIP code of the neighborhood where I grew up. Once I was finally logged in to my new account, it showed me a lost cat ad, which just so happened to belong to a woman who was my best friend in third grade. (She has a distinctive name, so I recognized it immediately.) Alongside a photo of her cat, the app displayed a map with a pin showing her location. We haven’t kept in touch over the years, but from this app I learned that she still lives within a mile of our elementary school. Were I up to no good, it’d be very easy to match the map’s pin with a house on Google Maps.

In this case, this old friend’s cat’s face has little to do with the privacy oversight, but nonetheless, by posting a photo of her cat in hopes that facial recognition will help her find him, she’s revealed quite a bit about herself. Other users could get hold of this information easily, just as I did, and in a way, that’s what you’re hoping will happen: that someone will see your lost pet ad and contact you, and you’ll be reunited with Rover.

But as with most apps, you may also be giving up a lot of data beyond what other users can see. Profile information and tracking data can reveal quite a bit about users to app developers and anyone the app is working with. In Finding Rover’s privacy policy, for instance, the company says it will collect information from you via cookies and web beacons, get info about you from third parties like social media platforms, and share information that “may be used by third parties for their own purposes,” which “may include offers for products or services that may interest you.”

That’s a no-brainer from a business perspective. If you had access to users’ location and details about the age, gender, and breed of their pets, targeted ads—pet food, toys, services—could be much more effective. “The default now is that people should just assume companies are selling their data to third parties, primarily for advertising,” says Jessica Vitak, an associate professor in the University of Maryland’s College of Information Studies. “Data has a financial value, so it’s in companies’ best interest that if they have the ability to collect the data, then they collect the data, even if they don’t have an immediate or direct use for it.”

After all, apps selling data to third parties is nothing new, and it’s how many free apps profit. “A lot of apps that you don’t pay for with money are making money based on your data, and you would never know this unless—and maybe even if—you read their privacy policy,” says Casey Oppenheim, CEO at Disconnect, a privacy app.

Polimeno says Finding Rover doesn’t sell data to marketers, and he sounded surprised when I mentioned that parts of the company’s privacy policy say they could give data to third parties. He explained that it’s not the company’s intention to do so, but that in the future, it may adopt in-app advertising for revenue. “[The app] will always be free, but there may be an ad in Finding Rover for dog food,” he says.

Vitak notes that it’s interesting to find an app that says it does not give data to third-party advertisers yet reserves the right to do so in the future. “A company that includes [those terms for data] but isn’t using it doesn’t make a lot of sense,” she says. Either way, that leaves the door open for future uses that are heretofore unknown, something many apps do, and most people don’t seem to care or notice. “This is one example of the normalization of the data collection process,” says Vitak, noting how difficult it is to convince people to think about the long-term ramifications. “How do you convey to consumers that a risk that is in the future, that is abstract and isn’t clearly defined, is important to think about, while they’re just focusing on the immediate benefit?” Providing information to save lost pets seems harmless now, but that data could be used for less innocuous purposes in the future.

Though Finding Rover may have good intentions for the data in its app, other apps and technologies that collect animal face recognition data might use it in more concerning ways. The Chinese version of Finding Rover by Megvii is also matching lost pets to owners by their faces—specifically, their nose biometrics—but hinted at the possibility of providing that data to the government as well so that pet owners who fail to leash their dogs or pick up poop are fined. Facial recognition is also being used by Chinese insurance company Ping An to enroll pets in health insurance. Given human health insurance companies’ use of people’s social media posts and fitness tracker data, it doesn’t seem so farfetched to imagine situations in which a dog’s daily walks, captured by local camera monitors, would garner lower premiums, or the dog’s Instagram videos could be used to argue against a filed claim.

All of this data is being stockpiled somewhere, dormant for who knows what later on. “It’s so easy to store this data virtually, potentially forever, that can be used for generations to come. What does this mean seven generations from now?” asks Oppenheim. What kind of pets you have is just one more piece of the puzzle that can be supplemented with other data your web presence has generated, creating a detailed profile of you. “Who knows what kind of world we’ll be living in where people have generations’ worth of data,” says Oppenheim. “The genie is out of the bottle and getting it back in is very, very difficult.”

Future Tense is a partnership of Slate, New America, and Arizona State University that examines emerging technologies, public policy, and society.