Meta Uses AI Bone Analysis to Detect Underage Users

Meta Deploys AI Body Analysis to Remove Underage Accounts

Meta is rolling out a new artificial intelligence system across Facebook and Instagram that analyzes users' photos and videos for physical characteristics — including height, bone structure, and skeletal proportions — to identify children under 13 and remove their accounts. The move represents one of the most aggressive and technically sophisticated approaches any major social media platform has taken to enforce age restrictions on its services.

The system marks a significant departure from traditional age-verification methods, which have largely relied on self-reported birthdates that minors can easily falsify. By leveraging computer vision and biometric analysis, Meta is essentially turning every photo a user uploads into a potential age-verification checkpoint.

Key Facts at a Glance

• Meta's new AI analyzes height, skeletal structure, and bone proportions in user-uploaded photos and videos
• The system targets accounts belonging to children under 13, the minimum age required by both Facebook and Instagram
• The technology goes beyond traditional self-reported age verification methods
• Both Facebook and Instagram will implement the new detection mechanism
• The initiative is part of Meta's broader push to improve child safety across its platforms
• The approach uses visual biometric markers that are difficult for young users to circumvent

How the AI Skeletal Analysis System Works

The technology behind Meta's new system draws on advances in computer vision and pose estimation — a field of AI research that has matured rapidly over the past 5 years. Modern pose estimation models can identify key anatomical landmarks in photographs with remarkable accuracy, mapping out the positions of joints, limbs, and the overall proportions of a person's body.

Children's skeletal structures differ from those of adults and adolescents in measurable ways. The ratio of head size to body length, limb proportions, shoulder width relative to height, and the development stage of certain bone structures all provide reliable indicators of a person's approximate age range. Meta's AI appears to leverage these anthropometric markers to flag accounts that likely belong to users under 13.

Unlike facial recognition — which has drawn intense scrutiny from privacy advocates and regulators — skeletal and body proportion analysis operates on a different level. It focuses on structural geometry rather than identifying who a person is, which may help Meta navigate the complex regulatory landscape surrounding biometric data.

Why Traditional Age Verification Has Failed

Social media platforms have struggled with underage users for well over a decade. The Children's Online Privacy Protection Act (COPPA) in the United States requires parental consent for the collection of personal information from children under 13, which is why most platforms set 13 as their minimum age requirement. However, enforcement has been notoriously weak.

The primary failure points of existing systems include:

• Self-reported birthdates are trivially easy to falsify — children simply enter a fake year of birth
• Parental consent mechanisms are rarely robust enough to confirm a parent's actual identity
• ID-based verification raises massive privacy concerns and creates friction that drives users away
• Credit card verification excludes users in regions where card penetration is low and doesn't guarantee the cardholder is the account user
• Knowledge-based quizzes designed to distinguish adults from children have proven unreliable

Meta's AI-driven approach sidesteps many of these issues by passively analyzing content that users voluntarily upload, rather than requiring them to complete an explicit verification step. This 'ambient verification' model could prove far more effective because it operates continuously rather than as a one-time gate.

The Technical Challenges and Limitations

While the concept is promising, several technical hurdles could limit the system's effectiveness. Age estimation from physical characteristics is inherently imprecise, particularly for children near the 13-year-old threshold. A tall 11-year-old and a short 14-year-old may have similar body proportions, creating a gray zone where false positives and false negatives are inevitable.

The system also raises questions about how it handles diverse body types across different ethnicities and populations. Growth rates and skeletal development timelines vary significantly across demographic groups, meaning the AI must be trained on an exceptionally diverse dataset to avoid systematic bias. If the model is disproportionately trained on data from one population, it could over- or under-flag children from other groups.

Another limitation is that the system presumably requires users to post photos or videos of themselves. Accounts that exclusively share memes, screenshots, or other non-personal content would be harder to evaluate. Sophisticated underage users who are aware of the system could theoretically avoid posting personal photos, though this would significantly limit their social media experience.

There is also the question of processing scale. Meta's platforms host billions of images. Running skeletal analysis on every uploaded photo in near real-time demands enormous computational resources, even for a company with Meta's infrastructure.

Privacy Concerns and Regulatory Implications

The deployment of AI body analysis inevitably raises privacy alarms. Scanning users' photos for physical characteristics sits in a legally complex space, particularly under regulations like the European Union's General Data Protection Regulation (GDPR) and various U.S. state-level biometric privacy laws such as Illinois' Biometric Information Privacy Act (BIPA).

Meta will likely argue that skeletal proportion analysis differs from traditional biometric identification. The system is not designed to identify individuals but rather to estimate age ranges — a distinction that could be legally significant. However, privacy advocates may counter that any systematic analysis of physical characteristics constitutes biometric data processing, regardless of the specific purpose.

Compared to Apple's controversial CSAM scanning proposal — which the company ultimately shelved in 2022 after widespread backlash over on-device photo scanning — Meta's approach targets a different problem but operates in a similarly sensitive space. The key difference is that Meta's system analyzes structural body proportions rather than image content matching, which may be perceived as less invasive.

Regulators in the EU, UK, and Australia have all been ramping up pressure on social media companies to better protect minors. The UK's Online Safety Act, which took effect in late 2023, imposes strict obligations on platforms to prevent children from accessing harmful content. Australia has gone even further, passing legislation to ban social media use for children under 16. Meta's proactive move could be seen as an attempt to get ahead of increasingly stringent regulatory demands.

Industry Context: A Growing Arms Race in Child Safety Tech

Meta is not the only tech giant investing in AI-powered age verification. Yoti, a London-based digital identity company, has developed facial age estimation technology that is already used by several platforms and retailers. Google has experimented with AI-based age estimation for YouTube, and TikTok has implemented its own detection systems to identify and remove underage accounts.

However, Meta's skeletal analysis approach appears to be unique in its methodology. Most existing systems rely on facial analysis — examining features like skin texture, facial structure, and other visual cues associated with aging. By focusing on body proportions and skeletal markers, Meta may achieve greater accuracy for the specific task of distinguishing pre-teens from teenagers, an age range where facial analysis tends to be less reliable.

The broader child safety technology market has grown substantially, with estimates suggesting it could reach $5 billion by 2027. This growth is driven not only by regulatory pressure but also by increasing public awareness and advertiser sensitivity around brand safety issues related to minors on social platforms.

What This Means for Users, Parents, and Advertisers

For parents, Meta's new system offers a potentially meaningful layer of protection. While no technology is foolproof, an AI system that continuously monitors for signs of underage users is far more robust than a one-time age gate during account registration. Parents who have struggled to keep younger children off social media may find some reassurance in the new approach.

For advertisers, the move is broadly positive. Brands have faced growing reputational risks associated with their ads appearing alongside content involving minors. A more effective system for removing underage accounts reduces this risk and strengthens Meta's pitch to advertisers who are increasingly concerned about brand safety.

For adult users, the practical impact should be minimal. The system operates passively in the background, analyzing content that is already being uploaded. However, users who are concerned about any form of automated body analysis may find the feature unsettling, even if it is designed with protective intent.

Looking Ahead: The Future of AI-Powered Age Verification

Meta's skeletal analysis system is likely just the beginning. As AI capabilities continue to advance, we can expect even more sophisticated approaches to age verification — potentially combining multiple signals including voice analysis, behavioral patterns, writing style, and content preferences alongside visual analysis.

The success or failure of Meta's system will be closely watched by regulators and competitors alike. If it proves effective at reducing the number of underage accounts without generating excessive false positives or privacy backlash, other platforms will face pressure to adopt similar measures. If it falters — either through inaccuracy or controversy — it could set back the broader effort to develop technological solutions for child safety.

One thing is clear: the era of relying on users to honestly self-report their age is coming to an end. The question is no longer whether platforms will use AI to verify age, but how aggressively and how accurately they can do so while respecting user privacy. Meta has placed a bold bet that skeletal analysis is part of the answer.