Meta’s AI to spot teens lying about age on Instagram

Under mounting scrutiny from parents and lawmakers over teen mental health concerns, Meta is deploying artificial intelligence to detect underage users lying about their age on Instagram, reportedly planning to automatically place suspected under-18 users into restrictive privacy settings. The system called the ‘adult classifier,’ will sift through data from users’ profiles, follower lists, and even ‘happy birthday’ messages from friends to help predict age.

Those found to be under 18 will reportedly be switched to new ‘teen accounts,’ introduced in September, which include tighter privacy controls such as limits on messaging and content visibility. Teens who attempt to increase their listed age will need to submit formal identification or a video selfie verified by the third-party service Yoti. The system, however, still allows for those aged 16 or 17 to manually adjust settings, while younger users need parental consent to turn off restrictions. 

These new measures come as the company faces lawsuit from several state attorneys general, accusing the company of contributing to a mental health crisis among teens. The company has also been challenged by a history of complaints, including whistleblower allegations in 2021 that Meta’s own research suggested Instagram could negatively impact teenage girls’ mental health. 

Enforcing age limits, however, remains a challenge. A recent study by the UK’s telecommunications regulator revealed that a third of minors with social media accounts claim to be 18 or older, and online age verification remains minimal on many platforms. Unlike other sites, Meta reportedly plans to flag accounts using the same email but with different birthdays and to check device IDs to identify users creating multiple accounts.

The age verification debate has led some executives to suggest that app stores, such as Google Play and Apple’s App Store, could help regulate age limits. But representatives from both Apple and Google have expressed concerns, citing privacy and data minimisation issues. 

For now, Meta is still refining its approach. Although accuracy improvements are underway, users wrongly classified as teens may appeal to have restrictions removed in the future.