It will now be more difficult to lie about your age on Instagram! The social network has just implemented a facial recognition verification system based on artificial intelligence, in the United States for the moment.
Instagram tightens its control of the age of its American users. While it has always been necessary to be over 13 to register and use the platform, no verification was required until 2019. Since then, the Meta group – which owns Facebook, WhatsApp and Instagram – has been using the platform. artificial intelligence to recognize the age of its users, in particular by detecting messages of the type “Happy 18e anniversary !or by comparing birthdates on Facebook and Instagram accounts – if they are linked. This helps prevent younger people from accessing Facebook Dating or receiving messages from unknown adults.
But Meta has decided to further strengthen these measures. Since June 23, 2022, when someone tries to change their date of birth on Instagram from under 18 to over 18, the platform asks for verification. Two means are offered to the user: go through three “guarantors” or submit to an artificial intelligence. The first method is a kind of guarantee: Instagram will ask the user to choose three of his contacts who will have to certify that he has passed his eighteenth birthday. Of course, they must also be 18 years old. The second solution is to submit to an algorithm. A demo version is available online, for those who wish to test it. This method is in place in the United States at the moment: There is no release date yet for other countries.
Facial recognition with AI to control age
The platform has worked in partnership with Yoti, whose technologies have been validated by the British government and German digital regulators. You have to send a photo to Instagram, which will then ask the algorithms to analyze the user’s face in order to offer an estimate of their age. But is it really reliable, when there are so many differences and distinct markers, like makeup, pimples or dark circles? According to one external study, Yoti’s tool is 98.89% reliable in determining whether an 18-year-old is over or under 25. But it is currently possible to fool the AI in other ways.
The journalists of The Verge were able to test this technology and discover its limits. It is possible to trick Yoti’s tool simply by showing a picture to the camera or asking someone older to take a picture of themselves. There is also the question of data protection, even if the two groups claim to delete the selfie afterwards and that the technology cannot recognize an identity – only detect age.