Meta will use AI to analyze height and bone structure in photos and videos to identify users under 13 on Facebook and Instagram, the company said in a Tuesday blog post. The system arrives weeks after a New Mexico jury ordered Meta to pay $375 million in civil penalties for misleading consumers about platform safety and putting children at risk. Meta has since threatened to pull its services from the state.
The visual analysis is live in a small set of countries and will expand from there. Meta is also rolling stricter Instagram Teen Accounts out to 27 countries across the EU and Brazil, and bringing Teen Accounts to Facebook in the U.S. for the first time, with the U.K. and EU following in June. Teen Accounts restrict DMs to existing connections, hide harmful comments, and default profiles to private.
Meta is careful about what the system is and isn't. "We want to be clear: this is not facial recognition," the company wrote. "Our AI looks at general themes and visual cues, for example height or bone structure, to estimate someone's general age; it does not identify the specific person in the image. By combining these visual insights with our analysis of text and interactions, we can significantly increase the number of underage accounts we identify and remove."
Key facts
- 01Meta's new AI analyzes height and bone structure in photos and videos to flag users it suspects are under 13.
- 02A New Mexico jury recently ordered Meta to pay $375 million for misleading consumers about platform safety.
- 03Meta is expanding Instagram Teen Accounts to 27 countries across the EU and Brazil.
- 04Facebook Teen Accounts launch in the U.S. now, with the U.K. and EU following in June.
- 05Flagged accounts are deactivated and require Meta's age verification process to avoid deletion.
The visual layer sits on top of existing signals Meta already mines for age cues — birthday posts, references to school grades, the language used in bios and captions. The model looks across posts, comments, captions, and profile metadata for a composite read. Meta said it plans to extend the system to Instagram Live and Facebook Groups.
“Our AI looks at general themes and visual cues, for example height or bone structure, to estimate someone's general age; it does not identify the specific person in the image.”— Jaeden Schafer
If the system flags an account as likely underage, Meta deactivates it. The user then has to clear Meta's age verification process to recover access; otherwise the account is deleted. That puts the burden of proof on the user, which is the inverse of how most platform moderation has historically worked.
The legal pressure behind this announcement is hard to miss. The New Mexico verdict didn't just deliver $375 million in penalties — it also ordered Meta to make structural changes to its platforms. Meta's response, threatening to exit the state, signals how aggressively the company is pushing back on state-level child safety mandates while simultaneously building the technical infrastructure those mandates effectively require.
Estimating a child's age from skeletal proportions is a real forensic technique, but applying it through consumer AI on user-uploaded media raises questions Meta's blog post does not answer. The company hasn't disclosed accuracy rates, false-positive rates, or how the model performs across body types, ethnicities, and disabilities. A 13-year-old who is tall for their age and a small adult occupy overlapping territory in any height-based classifier.
Meta's framing — that the model reads "general themes" rather than identifying individuals — is a deliberate distinction from facial recognition, which faces tighter regulation in the EU and several U.S. states. Whether regulators agree that bone-structure inference falls outside the facial recognition perimeter is a separate fight the company has now invited.
The skeptics' case writes itself. A user incorrectly flagged as underage has to surrender additional identity documents to Meta to keep their account, expanding the data the company holds on adult users in the name of protecting children. Meta has not said how long verification documents are retained or how the visual age model itself was trained.
Meta is in the position of every platform whose growth depended on being permissive about who could sign up: now that regulators and juries are pricing in the cost of underage users, the cheapest fix is more inference, more verification, more friction. Bone structure is just the latest signal in a stack that already reads everything you post. Whether that hardens trust or accelerates the next lawsuit depends on numbers Meta hasn't yet been willing to publish.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
