TL;DR: Yes, AI models have been training on public data for years, but 2025 brings new legal "teeth" to help parents claw that data back. The short version: stop posting "public" photos on Instagram, check the "AI Training" toggle in Snapchat settings, and teach your kids that every "hey ChatGPT" query is a permanent record.
Check out our guide on AI safety for kids
Ask our chatbot about the best privacy-first apps for 2026
If you feel like the internet suddenly got a lot creepier in the last year, you aren't imagining it. Generative AI models—the engines behind things like ChatGPT, Claude, and image generators like Midjourney—require trillions of data points to "learn" how to talk and create.
For a long time, tech companies treated the internet like a free buffet. If a photo of your kid at the park was public on Facebook, it was likely scraped and used to teach an AI what a "happy child" looks like. If your teen spent hours chatting with a bot on Character.ai, that conversation helped the bot get better at mimicking human emotion.
In 2026, we’ve reached a tipping point. We’re moving from the "Wild West" era of AI training into a "Regulated Era." New privacy rules now give parents more leverage, but the tech companies aren't going to just hand it to you—you have to know where to click.
When we talk about "personal data," we aren't just talking about a home address. We’re talking about:
- Biometric Data: Your child’s face and voice.
- Behavioral Data: What they find funny, what they're scared of, and how they talk (yes, even the "Ohio" and "Skibidi" slang).
- Intellectual Property: The art they draw on Procreate or the stories they write in Google Docs.
Once this data is "ingested" by an AI model, it’s incredibly hard to "un-learn." It becomes part of the machine's DNA. The concern isn't just a privacy leak today; it's how this data could be used to create deepfakes or targeted marketing profiles when they are adults.
As of 2026, several major updates to digital privacy laws (including the long-awaited COPPA 2.0 updates) have changed the game for parents:
- The Right to Opt-Out: Most major platforms are now legally required to provide a clear "Opt-Out of AI Training" toggle for users under 18.
- Voice Protection: It is now significantly harder for companies to store and use "voice prints" from apps like TikTok or YouTube Kids without explicit, verifiable parental consent.
- The "Data Delete" Mandate: Parents can now request that AI companies purge their child’s specific data from training sets—though "un-training" a model is still a technical nightmare, they are legally obligated to try.
Learn more about understanding COPPA 2.0
Meta has been very open (and a bit aggressive) about using public posts to train Meta AI. If your profile is public, your kids' faces are training fodder.
- The Fix: Set accounts to Private. Meta generally does not scrape private accounts for AI training (for now), but it’s better to be safe.
Snapchat (My AI)
Snapchat was one of the first to bake AI directly into the chat list. Every interaction your kid has with "My AI" is recorded and used to refine the experience.
- The Fix: Go into Settings > Privacy Controls > Clear My AI Data. Also, remind your kids that "My AI" is not their friend—it’s a data collector.
This is a massive hit with middle schoolers who want to "talk" to their favorite anime characters or historical figures. It is a data goldmine.
- The No-BS Take: This app is basically a giant vacuum for your kid’s personality. If they are using it, they are essentially volunteering to be a test subject for emotional AI.
Roblox is integrating more AI to help kids build games. While this is cool for "entrepreneurship," the platform is also monitoring chat and creation patterns to build its own proprietary models.
- The Fix: Check the parental controls on Roblox to limit who can interact with your child.
If you want to let your kids explore the digital world without feeling like they're being watched by a robot, consider these "cleaner" options:
DuckDuckGo (Search Engine)
Unlike Google, DuckDuckGo doesn't track your search history to build a profile. It’s the easiest privacy win you can get.
Signal (Messaging)
If your teen is ready for a phone, Signal is the gold standard. End-to-end encryption means not even the company can see what’s being sent, which means no AI training on those texts.
Kiddle (Visual Search)
A great alternative to a standard Google Image search for younger kids, Kiddle filters out a lot of the junk and doesn't have the same aggressive data-harvesting practices.
Ages 5-8: The "Ghost" Phase
At this age, kids should essentially be "ghosts" to AI. There is zero reason for an AI model to have their name, voice, or face.
- Action: Use YouTube Kids in "Signed Out" mode if possible.
- Action: Avoid "smart toys" that connect to the internet and record voices.
Ages 9-12: The "Incognito" Phase
This is when they start wanting to use Minecraft or Roblox.
- Action: Use pseudonyms. Never use their real name or birth year in a username.
- Action: Talk about the "AI Mirror." Explain that when they talk to a bot, the bot is just reflecting data back at them.
Ages 13+: The "Manager" Phase
Teens are going to use AI for homework (let's be real) and entertainment.
Check out our guide on how to protect your child's digital footprint
You don't need to give a lecture on data architecture. Just keep it real.
The "Permanent Marker" Analogy: "Hey, you know how if you write on a whiteboard with a permanent marker, it never really comes off? AI is like that. Everything you tell it, every photo you give it, it keeps forever. So let’s just be careful about what we’re 'writing' with."
The "Free Pizza" Analogy: "If a website is free, it’s because you are the product. They are 'eating' your data to get smarter. It’s okay to use it sometimes, but don't give them the whole kitchen."
The AI companies want us to feel like data collection is inevitable. It's not. While we can't scrub every single byte of data from the internet, we can significantly reduce the "surface area" our kids expose to these models.
By using the new 2025 privacy toggles, opting for privacy-first apps like Signal, and keeping family photos off public feeds, you are giving your child a "digital clean slate" that will be incredibly valuable by the time they hit adulthood.
- Audit your social media: Make sure your Instagram and Facebook are set to private.
- Toggle off training: Open ChatGPT or Snapchat on your kid's phone and find the "Data Training" or "History" settings to turn them off.
- Take the Screenwise Survey: Understand how your family's AI usage compares to your community.
Take the Screenwise Digital Habits Survey
Learn more about the best AI-free creative apps for kids