Meta uses AI to estimate users' ages after US court settlement

While the French Constitutional Council has just revoked the law which would have imposed generalized age verification on social networks, deemed too intrusive for private life. In the United States, Meta is taking the opposite path: as part of an $18 billion agreement concluded with 48 American states, the company must strengthen its ability to identify minors' accounts, without making them take out their identity cards. His solution: artificial intelligence that guesses age based on visual and behavioral cues.

An estimate, not an age verification: the nuance that changes everything

Technically, Meta's AI doesn't check anything at all, it estimates. On Instagram, the system crosses “contextual clues” (posts about a birthday, connection times typical of a school schedule, the list of friends, what an account focuses on) with an analysis of the published photos.

Meta herself explains that she looks for features in the images such as “height or bone structure” to estimate a general age, without identifying the person precisely. Combined, this dual analysis promises to spot more minor accounts. Google has already been doing the same thing on YouTube since last year, relying on viewing history rather than photos.

Meta under legal constraint

The deal resolves lawsuits from state attorneys general who accused Meta of knowingly designing Instagram and Facebook to capture the attention of young people at the expense of their mental health. It imposes specific obligations: regular external audits, quantified objectives on the rate of false positives (minors wrongly classified as adults), and above all a cascading mechanism.

As soon as an account is identified as belonging to a user under 13 and excluded from the platform, Meta must also verify the age of all its friends. He is given one year to refine the system. However, the company itself is campaigning for another solution: to bring age verification by Apple and Google to the level of application stores rather than platform by platform. Google refuses, arguing that the measure would ignore computers shared with family and pre-installed applications.

Europe is trying the same legal battle, with the same mixed results

On this side of the Atlantic, the method changes but the problem remains the same. Australia led the way in December 2025 with a ban on social media for under-16s. A study by the British Medical Journal has since shown that 85% of affected minors continue to access it despite everything. France, Greece, the United Kingdom and Spain are preparing similar texts, but the French precedent shows how perilous the exercise is: by imposing proof of age on any user, minor or not, the law censored by the Constitutional Council came up against exactly the same objection raised by the Electronic Frontier Foundation against Meta's approach in the United States, namely that the data collected to prove age can, one day, leak or be misused.

Neither the American judicial approach nor the European legislative approach have therefore yet resolved the same equation: protecting minors without transforming each Internet user, child or adult, into a suspect required to prove their identity.