Instagram AI Age Detection Goes Global: What Meta Says

Meta announced on Monday, November 9, that it is expanding the AI system it uses to guess a user’s age to Instagram accounts worldwide. If the system decides an account probably belongs to a teenager, the account is moved into Teen Account protections, even when the person signed up with an adult birthday.

That’s the company’s version. The details around it are messier, and the outlets covering it don’t quite agree.

What Meta actually said

Meta’s newsroom post is short on mechanics. It says the company uses “AI technology to find people we suspect to be teens,” that it has been “encouraged by the results,” and that millions more teens will now get the built-in protections. Teen Accounts limit who can contact a teen, what content they see and how much time they spend in the app.

The post also says the technology is still rolling out on Facebook and that “similar technology” comes to Threads later this year. In the US, parents will see an in-feed Instagram video this month explaining how teens are placed in these accounts. In Australia, Meta has started using Apple’s age-range signal, which passes an age bracket to apps when a parent sets up a teen’s iPhone.

What it doesn’t say

The post never lists what the model looks at. TechCrunch, The Verge and Engadget filled that in, and their accounts differ.

Question	TechCrunch	The Verge	Engadget
Signals named	Birthday messages, school-grade mentions, height and bone structure in photos	Birthday messages, accounts and content a user engages with, height and bone structure	Same list at first, then a correction
Who the photo analysis applies to	Not limited	Not limited	Added later: used to detect under-13s, per Meta
Earlier rollout	Not detailed	US, EU and Brazil	Not detailed

Engadget updated its story at 2:12 p.m. ET on November 9 after Meta told it that the bone-structure and school-grade cues are used to find people under 13 and remove them. That’s a narrower claim than the one in the original coverage, and the TechCrunch and Verge versions I read don’t carry it. The Next Web reports the first rollout covered the US, UK, Canada and Australia, with Brazil and the 27 EU states added in May, which doesn’t match The Verge’s shorter list either.

The settlement figure is also loose. TechCrunch says Meta settled with nearly every US state and territory in August for “up to $18 billion.” The Verge says $17.1 billion. Meta’s own post only says it reached an agreement with 52 attorneys general and promised to get better at spotting accounts that may belong to children under 13.

Why Meta is pushing this now

Meta’s post argues that one verified age signal from the phone, shared with every app, would work better than each company checking on its own. It says it has long backed laws that put age verification and parental consent in the app store. In the same post it calls on TikTok and YouTube to adopt its practices.

My read: that’s lobbying as much as product news. Meta would rather have Apple and Google carry the age-checking job, and an age-estimation system it has already built gives it a bargaining chip. I can’t verify that motive, and Meta doesn’t state it that way.

My read: the harder issue is errors. The system guesses, and a guess will sometimes land a 19-year-old in a Teen Account or miss a 15-year-old. Meta hasn’t published error rates, and I couldn’t find any in the coverage. Without them, “encouraged by the results” is a company assessment I can’t check.

This also connects to the wider fight over kids and AI. I covered France’s push for pre-audits of AI that reaches children, and eight US states are already suing OpenAI, as in Eight states sued OpenAI over kids and ChatGPT.

What to watch: the US parent notifications due this month, and whether Meta says anything about accuracy when Threads gets the same system.

Instagram's AI age detection goes global, but Meta's own post never says what it looks at

What Meta actually said

What it doesn’t say

Why Meta is pushing this now

Sources