Instagram is using AI to find teens who have falsified their age and is reclassifying them into “Teen Accounts” settings. The Meta social platform is reaching out to parents to tell them about “the importance of their teens providing the correct ages online, and tips to check and confirm their teens’ ages together,” but letting them know they “don’t have to go it alone — we’re using AI to help.” Instagram launched Teen Accounts last year as a way to enroll users and provide built-in protections. Those under 16 need a parent or guardian’s permission to change the setting. The safeguards are on by default for young users, limiting who can make contact and filtering viewable content.

“We’ve enrolled at least 54 million teens into Teen Accounts globally so far, with 97 percent of teens ages 13-15 electing to remain in these protections, explains an Instagram blog post from Meta, which expanded the Teen Account experience to Facebook and Messenger two weeks ago.

In a Facebook tech post explaining the ways in which AI is used to vet accounts, Meta explained “correctly categorizing teens is important to help prevent them from accessing adult features, like Facebook Dating or Mentorship.”

“While Meta has been using AI for age detection for a while, it says employing the tech in this way is a ‘big change,’” writes Engadget, noting that while the company has taken pains to ensure the age-detection tech is accurate, in case the AI makes a mistake users can appeal the finding and “have the option to change their settings and stick with an adult account.”

Verification methods include “detecting happy birthday posts and receiving reports from other users,” reports TechCrunch, explaining that “one of the most important ways parents can make sure their teens are in protected accounts is to check if their account lists their correct birthday.”

Social Media Today reports that to develop AI age filters, Meta trained its models “on signals such as profile information, like when a person’s account was created and interactions with other profiles and content” and also “trained on location-specific data to ensure it’s factoring in local trends.”