Meta Deploys AI to Remove Under-13 Accounts

Meta Targets Underage Users With New AI Age Detection System

Meta is preparing to deploy a sophisticated AI-powered age detection system across Facebook and Instagram, designed to identify and remove accounts belonging to users under 13 years old. The system combines visual scanning and text analysis to estimate user ages — without relying on traditional facial recognition or collecting personal identity documents.

The move, first reported by tech outlet NeoWin on May 5, represents one of the most ambitious attempts by a major social media platform to tackle the long-standing problem of underage users bypassing age gates. For years, children have simply entered false birth dates to create accounts, rendering existing age verification systems essentially useless.

Key Takeaways

• Meta is building an AI system that uses dual verification methods — visual scanning and text analysis — to detect underage users
• The system analyzes photos, videos, and posted content to estimate age ranges
• Meta emphasizes this is not facial recognition — it does not identify specific individuals
• The target is users under 13 who lied about their age during registration
• The approach attempts to balance child safety with user privacy concerns
• Physical features like height and skeletal structure are used as age indicators

How the Dual AI Verification System Works

Meta's approach splits age detection into 2 complementary streams that work together to build a confidence score about a user's actual age. Neither method alone would be sufficient, but combined, they create a more robust detection framework.

Visual scanning forms the first pillar of the system. The AI analyzes photos and videos uploaded by users, looking for physical indicators such as height, bone structure, and other developmental markers that correlate with age ranges. Importantly, Meta has been careful to distinguish this from facial recognition technology. The system does not attempt to identify who someone is — it only estimates how old they might be.

Text analysis serves as the second verification layer. The AI scans user-generated content for age-revealing signals, including birthday posts, references to school grades, mentions of specific age-related milestones, and other contextual clues that might indicate a user is younger than their registered age suggests. When a user posts 'happy 12th birthday to me' or mentions being in 6th grade, the system flags these as potential indicators of underage status.

Why Traditional Age Verification Has Failed

The internet's approach to age verification has been fundamentally broken since its inception. Nearly every platform that requires users to be a certain age relies on a simple self-declaration model — a dropdown menu or date field where users enter their birth date. There is no verification, no cross-referencing, and no consequences for lying.

Studies have consistently shown that children as young as 8 or 9 routinely create social media accounts by entering false ages. A 2023 report from the U.S. Surgeon General highlighted the mental health risks of social media for young users and called for stronger age verification mechanisms across the industry. Similarly, the Children's Online Privacy Protection Act (COPPA) in the United States sets 13 as the minimum age for data collection, but enforcement has been notoriously difficult.

Previous solutions have all faced significant hurdles:

• ID-based verification raises serious privacy concerns and creates barriers for legitimate adult users
• Credit card verification excludes users without payment methods and still does not confirm age
• Parental consent mechanisms are easily circumvented when children use their parents' devices
• Self-declaration is the current standard and is effectively meaningless
• Third-party age estimation services have faced accuracy and bias criticisms

Meta's AI-driven approach attempts to thread the needle by using behavioral and visual signals rather than requiring users to submit sensitive identity documents.

The Privacy Balancing Act

One of the most significant aspects of Meta's announcement is its explicit framing around privacy. The company is acutely aware that any age verification system risks triggering backlash from users concerned about surveillance and data collection.

By emphasizing that the system does not perform facial recognition and does not collect identity information, Meta is attempting to preempt criticism from privacy advocates. The AI estimates age ranges rather than determining exact ages, and it analyzes content that users have already voluntarily shared on the platform.

However, privacy experts are likely to raise questions about this approach. Scanning user photos for physical characteristics and analyzing post content for age signals could be perceived as a form of surveillance, even if the stated purpose is child protection. The system essentially monitors user behavior and physical appearance continuously, which sits uncomfortably with growing global demands for less intrusive data practices.

There is also the question of accuracy and bias. AI age estimation systems have historically performed unevenly across different demographics. Research from organizations like the National Institute of Standards and Technology (NIST) has shown that facial analysis algorithms can exhibit varying accuracy rates based on race, gender, and ethnicity. While Meta says its system is not facial recognition per se, any visual analysis of human physical features carries similar risks of uneven performance.

How This Compares to Industry Efforts

Meta is not the only tech giant grappling with underage user detection. The broader industry has been moving toward AI-assisted age verification, driven by regulatory pressure and public concern.

YouTube has experimented with age estimation technology from companies like Yoti, which uses selfie-based analysis to estimate whether a user meets age requirements. TikTok has removed millions of accounts suspected of belonging to underage users, though its methods have been less transparent. Apple and Google have implemented parental control features at the operating system level, but these require active parental involvement.

Compared to these approaches, Meta's dual-method system stands out for its comprehensiveness:

• YouTube's Yoti integration requires users to actively submit a selfie for verification — Meta's system works passively on existing content
• TikTok's approach has relied more heavily on user reports and manual review — Meta is automating the detection pipeline
• OS-level parental controls from Apple and Google only work when parents configure them — Meta's system operates independently
• Legislative approaches like the UK's Online Safety Act and Australia's proposed social media bans take a regulatory angle — Meta's is a platform-level technical solution

The European Union's Digital Services Act (DSA) and the UK's Online Safety Act are both increasing pressure on platforms to protect minors, making proactive systems like Meta's not just a public relations move but a regulatory necessity.

What This Means for Users and Parents

For the average Facebook or Instagram user over 13, the immediate impact should be minimal. Meta has not indicated that adult users will need to take any additional verification steps. The system operates in the background, analyzing content that is already on the platform.

For parents, this represents a potentially significant development. Many parents have struggled to prevent younger children from accessing social media, and platform-level detection could serve as a safety net when parental controls fail or are not implemented.

However, several practical questions remain unanswered. Meta has not disclosed what happens when an account is flagged — whether the user receives a warning, whether there is an appeals process, or whether account removal is immediate and permanent. The company also has not shared accuracy benchmarks for the system or outlined how it will handle false positives, which could affect legitimate teenage users who appear younger than their actual age.

Looking Ahead: The Future of AI Age Verification

Meta's dual-method AI age detection system signals a broader industry shift toward using artificial intelligence for platform governance. If successful, it could set a precedent that other platforms adopt, creating a new standard for how social media companies handle underage users.

The timeline for full deployment remains unclear. Meta has not announced a specific launch date, and the system may undergo extended testing before rolling out globally. Regulatory environments vary significantly across markets — what is acceptable in the United States may face different scrutiny under European privacy laws like the General Data Protection Regulation (GDPR).

The long-term implications extend beyond just age verification. If Meta demonstrates that AI can effectively analyze user behavior and physical characteristics to determine age, the same technology framework could theoretically be applied to other verification challenges — raising both opportunities and concerns about the expanding role of AI in moderating online spaces.

For now, the key question is whether Meta's approach can achieve the accuracy and fairness needed to meaningfully protect children without creating new privacy risks or disproportionately affecting certain user groups. The tech industry, regulators, and parents alike will be watching closely as this system moves from concept to reality.