Real kids, internet fakes – how Facebook is using your photos to train AI, and what you can do to stop it

Your innocent family photos may be being used to train Facebook’s artificial intelligence app to create AI deepfakes. Here’s how to stop it happening.

Australian mums and dads are being urged to follow Hollywood stars and keep their children off public social media pages – or risk having their images harvested by the big tech giants.

A-list Hollywood couples Eva Mendes and Ryan Gosling, Mila Kunis and Ashton Kutcher, along with Aussie celebs Sam and Lara Worthington, and Keith Urban and Nicole Kidman, all keep their children out of the spotlight to protect their privacy and emotional wellbeing.

But now there’s another risky reason to keep Facebook pages private. Tech giant Meta has admitted harvesting images to train its AI app, including those of children posted by their parents on public pages.

An AI image of a child being rescued from floods that went viral on Facebook because many older users didn't realise it was fake.

Susan McLean, Australia’s cyber cop, is issuing a “reality check” for parents, urging them to protect their children by making sure their accounts are private.

“I don’t know why anyone would have a publicly accessible social media account. You would never have put half a dozen cute photos of your kid on a poster on your front fence so don’t do it on the internet,” says Ms McLean, a former Victoria Police officer who is now an expert in cyber safety and young people.

“So often we see people sign up for these, don’t read the Ts and Cs and then complain. The internet is not set and forget. You should directly check your account to make sure that they are doing what you are expecting them to do.

Australia’s “cyber cop” Susan McLean.

“Understand that if your account is set to public, then they can take the photos that you post and use it to train AI. It’s just another risk that occurs when you’re out in public on the internet.

“I think that there is still a reluctance of parents to fully understand the risk involved in posting large amounts of photos of their children to social media.”

Dr Kristyn Sommer took action earlier this year to remove her son, now two, and six-year-old daughter from her social media accounts.

The popular Queensland influencer and mum-of-two, who has nearly 500,000 followers on TikTok and 55,000 on Instagram, says she is still completing the mammoth task of erasing their digital footprint, which included one video of her daughter that had racked up 18m views.

Influencer Dr Kristyn Sommer, who has removed her two children from her social media platforms.

She says she took the bold move out of respect for their privacy as individuals who “have their own stories, despite the fact that our stories are intrinsically linked”.

“It is not my right to share their stories ... it was very clear that I should not do that,” says Dr Sommer.

“But I see lots and lots of people really wrestling with this because it is a social norm to share about your children online. It is very much like a social rebellion to not share your kid online.”

Dr Sommer, a research scientist who is helping to design robots to work with children, says the emerging technology added another problematic factor to the practice of “sharenting”.

“I’m just a parent who posted pictures of my kids because they’re gorgeous ... and I don’t know how I feel about AI using them. The fact that it (AI) uses the stories and images that you share about your own children to then learn more and to get better at what it does can have far-reaching impacts that we don’t really know the outcomes of,” she says.

“Even for someone like me who understands the way that AI works, I still find it really hard to grapple with future ramifications of sharing my children’s face and stories online ... so I’m just doing the black-and-white thing of just taking my kids offline. I’d rather make mistakes for myself instead of making mistakes for my children because that is not fair.

“In the end, you need to think about the way that you’re creating your child’s digital footprint and we should centre children’s rights to privacy and autonomy and independence and just the fact that they’re their own human at the centre of all of these decisions.”

Jenny Branch-Allen, chief executive of the Australian Parents’ Council, says it will be “really hard to shut the gates” on the practice of “sharenting”.

Ms Branch-Allen pointed to the Danish government’s revolutionary new law to tackle deep fakes that gives individuals copyright over their own image, including their face.

“We’ve lost control of our image so it can be used by anyone, really, at any time and we have to stop that because, honestly, the damage that it can do (to children) is insurmountable,” she says.

Jenny Branch-Allen says when you upload family photos to social media, you’ve lost control of them.

She says she knows of parents who are closing their Facebook pages and “becoming very wary about sharing these images” but they are “in the minority at the moment”.

“What was a great way to connect and we thought was a lovely way to share special moments has now become a monster that is really impacting on families,” she says.

“Who knew the ugly side of this was going to get so ugly?”

Meta has admitted to harvesting public Facebook data – including photographs and posts dating back to 2007 – and using it to train its free Meta AI app, which uses the data to create images that look like real-life photographs.

Facebook has a policy of not scraping the pages of children under the age of 18.

“Globally, generative AI models are trained using a large mix of data sources, including publicly available information online and licensed data,” says a Meta spokesperson.

“Using public information for training is common across the industry and not unique to Meta.”

Got a story tip for us? Email [e-mail]