Proton
Iscriviti con RSS
Family photos linked by AI, suggesting that your family photos may be used for training AI

Your family photos are training AI – here’s why it matters and how to stop it

Elena Constantinescu

Condividi questa pagina

The family photos you share online, such as weddings, birthdays, and everyday moments, can end up in places you never imagined: helping train AI models. These aren’t just stock photos or staged portraits. They’re real snapshots of your life, kids, partner, and memories — quietly pulled from social media, blogs, and even old public albums.

And without your knowledge or consent, Big Tech companies have used these images with your valuable data to build facial recognition systems(nuova finestra), train surveillance tools(nuova finestra), and generate synthetic scenes that feel uncomfortably familiar.

Some pressure is being placed on companies to limit data scraping and enforce consent. The EU’s new AI Act(nuova finestra), for example, bans certain high-risk uses and sets stricter rules for training data. But for now, most platforms treat public content as fair game.

While some opt-out options exist(nuova finestra), they’re buried, unclear, or rarely offered upfront. And even if you do get your images removed, the AI has already learned from them. That’s the catch: once your face is part of the model, it stays there.

So, what can you do to keep your family photos safe? While laws are still catching up, the practical steps in this guide can help you reclaim control and keep your cherished memories out of AI datasets.

How Big Tech uses family photos to train AI

Big Tech has been scrutinized(nuova finestra) for how it uses public images, including family photos, to train AI models. Here’s how that process often happens:

  • Scraping public images: Companies collect publicly available images(nuova finestra) from the internet, including family pictures posted on social media, blogs, or image hosting sites.
  • Datasets built from these photos: The photos are used to build large datasets, often labeled with metadata like age, gender, emotion, or activity — such as laughing, hugging or family gathering. These help train AI models to recognize human faces, expressions, and even relationships.
  • Use in facial recognition and surveillance: Family photos help train systems to identify people in various lighting, angles, and group contexts from datasets(nuova finestra) — making the tech more accurate in real-world conditions.
  • Training generative AI: Photos of families help these models understand human composition, realistic skin tones, clothing styles, and other factors. For instance, happy family at a picnic scenes rely on training data that included actual images of families having picnics.

Examples of public images scraped for AI

Here are some of the most well-known cases where personal images, including family photos, were used without consent:

  • MegaFace dataset: The University of Washington used photos from Flickr(nuova finestra), including family albums and personal moments, to create a facial recognition dataset. This dataset was used by Google, Amazon, Microsoft, other companies, and government agencies worldwide. People found their own or their children’s faces in the dataset without knowing.
  • Clearview AI: Billions of photos from Facebook, Instagram, LinkedIn, and other websites were scraped to build a powerful facial recognition tool used by police and government agencies. People found themselves, their kids, and even deceased relatives in police databases. Clearview AI has since faced lawsuits(nuova finestra), fines, and legal warnings for violating people’s privacy.
  • LAION-5B dataset: LLAION-5B, used to train AI models like Stable Diffusion, included images of children from personal blogs, social media, and stills from YouTube videos with small view counts. Human Rights Watch found over 170 images(nuova finestra) of Brazilian children used without consent.
  • Facebook’s DeepFace: This deep learning facial recognition system(nuova finestra) uses publicly available photos and can identify people in photos with over 97% accuracy — to compare, FBI’s technology has 85% accuracy.

What you risk when your memories train AI

  • Opt-outs exist, but only after the fact: Some platforms let you request image removal from datasets. But unless the model is retrained — which rarely happens and there’s no clear way to know(nuova finestra) — your data stays baked in.
  • Can’t control the narrative: Once AI learns from your photo, you no longer control how it’s used. Your face, smile, or the way you hug your child could live on in synthetic images used by others without your knowledge.
  • Invisible targeting: AI trained on your family photos can recognize what you look like, how you dress, and where you live. That data can sharpen ad targeting(nuova finestra) or make facial recognition systems more effective on you and your loved ones.
  • Surveillance without consent: Your image could end up in AI training for facial recognition systems used by law enforcement(nuova finestra) or other surveillance tools without your knowledge. These systems may later identify you in public spaces, airports, protests, or social platforms, based solely on photos you once shared.
  • Exploited for ransom: Bad actors can use AI to turn your photos into deepfakes — sometimes explicit or abusive. In one case, a 14-year-old was targeted with a deepfake image(nuova finestra) and blackmailed by a predator threatening to share it unless paid.

How to keep your family photos private

Here are practical steps you can take online:

Lock down your privacy settings

Limit the pictures you expose online

  • Be cautious about location tagging and image descriptions. That extra context you add when sharing photos online can reveal your home, school, work, favorite spots, who’s in the photo, or what they’re doing.
  • Remove metadata from photos, such as GPS coordinates or date and time. Doing this ensures AI systems scraping images have less personal data to work with, even if they access the photo itself.
  • Review app permissions to revoke photo library access to apps you rarely or never use.
  • Blur or mask faces — especially of children or vulnerable individuals — if you’re posting publicly. AI can still learn from altered images, but it makes scraping harder and reduces quality for training.
  • Avoid uploading to AI tools, such as apps that offer face filters, age changers, animated designs(nuova finestra), or AI family portrait generators. They often store and reuse your uploaded images.
  • Talk to family, friends, and schools about photo privacy. Ask them not to post identifiable pictures without permission, avoid adding personal context like names or locations, and encourage private sharing groups and clear consent policies at events and schools.

Check and remove images from public AI datasets

You can use Have I Been Trained(nuova finestra) to search for your pictures in datasets like LAION-5B. If found, request dataset removal or contact dataset creators directly. However, your photos may still exist in already-trained AI models.

Keep your family photos safe with Proton Drive

Proton Drive is secure cloud storage that offers true privacy by default. Unlike Google Photos, which has unclear AI policies, or iCloud, where end-to-end encryption (E2EE) is optional and no longer available in the UK, Proton Drive is safe by design for sharing and storing photos.

End-to-end encryption is automatic and always on — your photos, albums and their metadata are encrypted before they leave your device, and there’s no way for us or anyone else to see your files. Your pictures are never used for AI training, and they stay private, no matter where you are.

When it comes to sharing, Proton Drive makes it easy to send photos to relatives safely using email invites or expiring, password-protected links. A dedicated tab lets you manage and revoke shared access at any time.

Your memories deserve more than trust — they deserve proof. Proton Drive is open source and independently audited, so your family photos stay safe, and you stay in control.

Mantieni privati i tuoi file e condividili in modo sicuro
Passa gratis a Proton Drive

Articoli correlati

Microsoft has announced that, starting in June, you’ll no longer be able to save new passwords in the Microsoft Authenticator app.
en
If you want to use Microsoft’s password management features, you now need to step deeper into Microsoft’s walled garden. There's another way.
An illustration of a photo containing a parent and their child, overlayed on a cloud and an open padlock
en
Is your family’s photo collection safe? We surveyed 2,000 UK parents to uncover the truth about cloud storage risks, data breaches, and protecting precious memories.
en
Albums in Proton Drive makes it easier than ever to store and manage photos and videos while protecting your memories with end-to-end encryption.
What is SSO and why is it useful for businesses?
en
SSO allows employees to securely access all their work apps with just one login. Here's how SSO works and why businesses should consider it.
adolescence and the internet we handed to kids
en
Adolescence shows how platforms shape kids in harmful ways. Here's why transparent, open digital education must lead the change.
file management for teams
en
Learn how to set up a secure, efficient file management system for your team and keep everything organized with Proton Drive.