Monday, October 7

Can You Hide a Child’s Face From A.I.?

There are two distinct factions of parents on TikTok: those who will crack eggs over their kids’ heads for likes and those who are trying desperately to make sure the internet doesn’t know who their children are.

For the 35-year-old TikTok star who posts under the name Kodye Elyse, an uncomfortable online experience made her stop including her three children on her social media. A video she posted in 2020 of her young daughter dancing attracted millions of views and creepy comments from strange men. (She requested that The New York Times not print her full name because she and her children have been doxxed in the past.)

“It’s kind of like ‘The Truman Show’ on the internet,” said Kodye Elyse, who has four million followers on TikTok and posts about her work as a cosmetic tattoo artist and her experiences as a single mother. “You never know who’s looking.”

After that experience, she scrubbed her children’s images from the internet. She tracked down all of her online accounts, on sites such as Facebook and Pinterest, and deleted them or made them private. She has since joined the clamorous camp of TikTokers encouraging fellow parents not to post about their children publicly.

But in September, she discovered her efforts hadn’t been entirely successful. Kodye Elyse used PimEyes, a startling search engine that finds photos of a person on the internet within seconds using facial recognition technology. When she uploaded a photo of her 7-year-old son, the results included an image of him she had never seen before. She needed a $29.99 subscription to see where the image had come from.

Her ex-husband had taken their son to a soccer game, and they were in the background of a photograph on a sports news site, sitting in the front row behind the goal. She realized she wouldn’t be able to get the news organization to take down the photo, but she submitted a removal request, via an online form, to PimEyes, so that her son’s image would not show up if other people searched for his face.

She also found a toddler-aged photo of her now 9-year-old daughter being used to promote a summer camp she had attended. She asked the camp to take down the photo, which it did.

“I think everybody should be checking that,” Kodye Elyse said. “It’s a good way to know that no one is repurposing your kids’ images.”

How much parents should post about their children online has been discussed and scrutinized to such an intense degree that it has its own off-putting portmanteau: “sharenting.”

Historically, the main criticism of parents who overshare online has been the invasion of their progeny’s privacy, but advances in artificial intelligence-based technologies present new ways for bad actors to misappropriate online content of children.

Among the novel risks are scams featuring deepfake technology that mimic children’s voices and the possibility that a stranger could learn a child’s name and address from just a search of their photo.

Amanda Lenhart, the head of research at Common Sense Media, a nonprofit that offers media advice to parents, pointed to a recent public service campaign from Deutsche Telekom that urged more careful sharing of children’s data. The video featured an actress portraying a 9-year-old named Ella, whose fictional parents were indiscreet about posting photos and videos of her online. Deepfake technology generated a digitally aged version of Ella who admonishes her fictional parents, telling them that her identity has been stolen, her voice has been duplicated to trick them into thinking she’s been kidnapped and a nude photo of her childhood self has been exploited.

Ms. Lenhart called the video “heavy-handed” but said it made the point that “actually this technology is really quite good.” People are already receiving calls from scammers imitating loved ones in peril using versions of their voices created with A.I. tools.

Jennifer DeStefano, a mother in Arizona, got a call this year from someone who claimed to have kidnapped her 15-year-old daughter. “I answered the phone ‘Hello;’ on the other end was our daughter Briana sobbing and crying saying, ‘Mom,’” Ms. Stefano said in congressional testimony this summer.

She was negotiating to pay the kidnappers $50,000 when she discovered her daughter was at home “resting safely in bed.”

Obscure online photos and videos might be linked to someone’s face with facial recognition technology, which has grown in power and accuracy in recent years. Photos taken at a school, a day care, a birthday party or a playground could show up in such a search. (A school or day care should present you with a waiver; feel free to say no.)

“When a child is younger, the parent has more control over their image,” said Debbie Reynolds, a data privacy and emerging technologies consultant. “But kids grow up. They have friends. They go to parties. Schools take pictures.”

Ms. Reynolds recommends that parents search online for their children’s faces using a service like PimEyes or FaceCheck.ID. If they don’t like what comes up, they should try to get the websites the photo was posted on to take it down, she said. (Some will, but others — like news organizations — might not.)

In a 2020 Pew Research survey, more than 80 percent of parents reported sharing photos, videos and information about their children on social media sites. Experts were unable to say how many parents are sharing those images only on private social media accounts, as opposed to publicly, but they said that private sharing is an increasingly common practice.

When I share digital photos of my daughters, I tend to use private messaging apps and an Instagram account limited to friends and family. But when I searched for their faces on PimEyes, I also discovered a public photo I had forgotten about — that accompanied a story I had written — of my now 6-year-old daughter when she was 2. I requested that PimEyes remove the image from its results, and it no longer appears in a search.

While a public face search engine is a potentially useful tool for a parent, it could also be used nefariously.

“A tool like PimEyes can be — and likely is — used as easily by a stalker as it is a concerned parent,” said Bill Fitzgerald, a privacy researcher, who also expressed concern about overbearing parents using it to monitor their teen children’s activities.

PimEyes’ owner, Giorgi Gobronidze, said more than 200 accounts had been deactivated on the site for inappropriate searches of children’s faces.

A similar face recognition engine, Clearview AI, whose use is limited to law enforcement, has been used to identify victims in photos of child sexual abuse. Mr. Gobronidze said PimEyes had been used similarly by human rights organizations to help children. But he is worried enough about potential child predators using the service that PimEyes is working on a feature to block searches of faces that appear to belong to minors. (Mr. Fitzgerald, the privacy researcher, is concerned that parents using the tool to look for their own children, might be unintentionally helping the PimEyes algorithm improve its recognition of those minors.)

Mimi Ito, a cultural anthropologist and director of the Connected Learning Lab at the University of California, Irvine, said facial recognition technology makes the otherwise joyful sharing of children’s photos online more challenging.

“There’s a growing awareness that with A.I., we don’t really have control of all the data that we’re spewing into the social media ecosystem,” she said.

Lucy and Mike Fitzgerald, professional ballroom dancers in St. Louis who maintain an active social media presence to advertise their business, refrain from posting images of their daughters, ages 5 and 3, online, and have asked friends and family members to respect the prohibition. They believe their daughters should have the right to create and control their own online footprints. They also worry their images might be used inappropriately.

“The fact that you can steal someone’s photo in a couple of clicks and then use it for whatever you want is concerning,” Ms. Fitzgerald said. “I understand the appeal of posting your kids’ photos, but ultimately, we don’t want them to be the ones to have to deal with potential unintended consequences.”

Ms. Fitzgerald and her husband are not experts who were “informed about what’s looming on the horizon of tech,” she said. But, she added, they “had a feeling” years ago that there were “going to be capabilities that we can’t foresee right now that will eventually be problematic for our kids.”

Parents more likely to know specifics about what’s looming on the tech horizon, including Edward Snowden, the National Security Agency contractor turned whistle-blower, and Mark Zuckerberg, the Facebook co-founder, conceal their children’s faces in otherwise public social media posts. In holiday-themed posts on Instagram, Mr. Zuckerberg used the clumsy emoji method — posting a digital sticker on his older children’s heads — while Mr. Snowden and his wife, Lindsay Mills, artfully posed one of their two sons behind a balloon to obscure his face.

“I want my kids to have the option to disclose themselves into the world, in whatever form they choose, whenever they are ready,” Ms. Mills said.

A spokeswoman for Mr. Zuckerberg declined to comment, or to explain why his baby’s face didn’t get the same treatment, and whether it was because facial recognition technology doesn’t work very well on infants.

Many experts noted that teens think a lot about how they curate their digital identities, and that some use pseudonyms online to prevent parents, teachers and potential employers from finding their accounts. But if there is a public image on that account that features their face, it could still be linked back to them with a face search engine.

“Your face is very hard to keep off of the web,” said Priya Kumar, an assistant professor at Pennsylvania State University who has studied the privacy implications of sharenting.

Dr. Kumar suggests that parents involve children, around the age of 4, in the process of posting — and talk to them about which images are OK to share.

Amy Webb, the chief executive of Future Today Institute, a business consultancy that focuses on technology, pledged in a Slate post a decade ago not to post personal photos or identifying information of her toddler online. (Some readers took this as a challenge, and found a family photo Ms. Webb had inadvertently made public, illustrating just how hard it can be to keep a child off the internet.) Her daughter, now a teenager, said she appreciated being an “online ghost,” and thought it would help her professionally.

Future employers “are going to find literally nothing on me because I don’t have any platforms,” she said. “It’s going to help me succeed in my future.”

Other young people who have grown up in the age of online sharing said they too were thankful to have parents who did not post photos of them publicly online. Shreya Nallamothu, 16, is a high school student whose research on child influencers helped lead to a new Illinois state law that requires parents to set aside earnings for their children if they are featuring them in monetized online content. She said she was “very grateful” that her parents didn’t post “super embarrassing moments of me on social media.”

“There are people in my grade who are really good at finding your classmates’ parents’ Facebook and scrolling down,” she said. They use any cringeworthy fodder for disappearing birthday posts on Snapchat.

Arielle Geismar, 22, a college student and digital safety advocate in Washington, D.C., described it as a “privilege to grow up without a digital identity being made for you.”

“Kids are currently technology’s guinea pigs,” Ms. Geismar said. “It’s our responsibility to take care of them.”