Sharenting in the AI Age: Your Kid's Face Is Already Training Data

Kids & AI / Privacy

A few years ago, the worry with sharenting was embarrassment. Would the class see the bath photo? Would a future employer scroll back? Those worries haven't gone away. They've been joined by a quieter, stranger one. Every photo you post of your child is now potential raw material for a machine that will outlive the post.

I'm not saying don't take the picture. I'm saying know where the picture goes after you hit share.

Your kid's digital birth came before the real one

We like to think a child's online life starts when they pick up a device. It doesn't. A 2010 AVG study found that by age two, 92% of US children already have some kind of online presence, with the average digital birth landing around six months old. A third of US babies have sonogram photos online before they're born. Their first post beats their first breath.

The footprint is laid down by someone else, years before the child can say the word privacy. Consent was never on the table.

A toddler's smile is now a training sample

That pile of childhood images doesn't just sit there. It trains things.

Barclays projected that by 2030, sharenting could account for two-thirds of all online identity theft against young people, roughly 7.4 million incidents a year. The bank's point was simple. A first pet's name, a school badge, a birthday sign in the background, these are the answers to the security questions a fraudster needs. We post the cheat sheet without meaning to.

Identity theft is the slow-burn version. The faster version is the face.

When strangers can rebuild your child's face

In 2026, the University of Adelaide and ICMEC Australia ran the first population-level study of AI in child sexual victimisation. They found AI involved in about 26%, roughly one in four, of non-consensual sexual image sharing among under-18s.

The mechanism is a nudify service: it takes a real photo of a real child and builds something abusive from it. The source image is often innocent. A soccer photo. A school play. Something a parent proudly shared.

Nobody is saying your feed is the weapon. The point is that a child's likeness, once scattered across the open web, is no longer theirs to recall. The genie doesn't go back in the bottle, and now the genie can be cloned.

Most parents haven't had the AI talk

While we argue about whether to post the birthday cake, our kids have already moved on to the machines.

Australia's eSafety Commissioner surveyed 1,950 children aged 10 to 17 in 2026. 78% had used an AI assistant like ChatGPT, and one in five who used one reported a potentially harmful interaction, from controlling behaviour to sexual content. About a third had shared personal or sensitive information with the tool. The kids are already in the room with AI. Most parents are still in the hallway.

That disconnect is the real risk. The fact that we manage the old posts while the new ones are being typed by our children into a chatbot we've never opened.

Keep the camera off the feed

The answer isn't panic or deletion. It's architecture. Build the kid's digital life so the risky parts never touch the open web.

The parents who get this right aren't the ones who post nothing. They're the ones who treated consent as a habit. The consent habit starts earlier than you think. And the families learning to hand kids agency instead of tracking them see the same lesson from the other side.