What the Hidden NameTag System Is and How It Was Found
Meta smart glasses facial recognition refers to dormant code embedded in Meta’s AI glasses companion app that can detect faces, generate biometric identifiers, and match people using data stored locally on a user’s phone, raising smart glasses privacy concerns because this powerful capability was shipped without clear disclosure, consent flows, or user-facing controls. The unreleased feature, internally referred to as “NameTag,” was uncovered through reverse engineering by independent researchers examining the Meta AI app, which is required for products like the Ray‑Ban Meta smart glasses and has been installed more than 50 million times. Their technical analysis revealed three machine learning models working together to detect faces, align them, and create 2048‑dimensional biometric fingerprints. WIRED’s reporting highlighted that this system sat dormant in production builds, meaning millions of users unknowingly carried a complete facial recognition pipeline on their phones even though no public feature advertised it.
How Meta’s On-Device Facial Recognition Pipeline Worked
The face recognition code discovery showed an infrastructure closer to a surveillance platform than a casual camera enhancement. Researchers from Buchodi and EFF’s Threat Lab describe three ExecuTorch models—SCRFD for face detection, KPSAligner for positioning, and a scaled-up SFace variant for embedding generation—working in sequence to turn faces into biometric data tracking entries. These embeddings are stored in an SQLite vector database designed for similarity search, while unknown faces are saved as cropped images plus binary embedding files that persist across device reboots. The app also contained a notification system that could display “Person recognized” alerts and internal text for a hidden “Connections” feature promising to “remember the people you met.” In testing, the pipeline successfully identified a portrait of Michel Foucault after his biometric template was preloaded, proving the facial recognition surveillance chain was functional end-to-end despite remaining invisible to everyday users.
Meta’s Explanation, Past Faceprint Controversies, and Legal Baggage
Meta’s response has been to frame NameTag as internal experimentation rather than a near-term feature launch. Company spokesperson Ryan Daniels said the code reflects exploration and that Meta is not building a central face database, stressing that any consumer release would be careful and transparent. This assurance sits awkwardly beside Meta’s history: in 2021 the company announced it was shutting down Facebook’s facial recognition system and deleting more than a billion faceprints after years of criticism, regulatory scrutiny, and costly biometric privacy settlements in multiple states. Privacy advocates argue that shipping dormant facial recognition code to tens of millions of devices revives old concerns under a new form factor. According to EFF, Meta’s decision to remove the technology within about 48 hours of WIRED’s report shows that public scrutiny still carries weight, but it does not guarantee the company has abandoned similar ambitions for future products.
Why Wearable Facial Recognition Alarms Privacy Advocates
For critics, the core danger is that Meta smart glasses facial recognition could normalize ever-present biometric surveillance. Because smart glasses sit on a wearer’s face, they can quietly record people in public and feed those images into systems that generate persistent biometric identifiers without their knowledge. The hidden NameTag system shows how easy it is to turn AI wearables into always-on recognition tools that “remember” strangers, building what EFF calls a “distributed surveillance machine.” Even though Meta says the system runs on-device and not in a central database, transparency disclosures indicate that data from AI glasses can still move between phones and remote servers, widening the attack surface if such features went live. For many privacy advocates, this episode confirms that clear consent, visible controls, and strict limits on biometric data tracking must precede any deployment of facial recognition on consumer wearables.
Code Removal, Remaining Questions, and What Comes Next
After WIRED exposed the face recognition code discovery and public backlash mounted, Meta shipped a June 5 update to its AI app that stripped out the facial recognition models, the “Person recognized” notification logic, and the databases used to store biometric signatures. EFF calls this a victory, but stresses that “quiet deletion of code does not equal a permanent change of heart.” Meta has declined to answer whether NameTag might return in another form or what happened to any biometric data collected during internal tests. Regulators and civil society groups see the episode as a warning that companies can ship dormant surveillance features and activate or withdraw them at will. The incident is likely to fuel calls for stronger biometric privacy laws, explicit opt-in rules, and enforcement mechanisms so that the future of wearable AI is not defined by hidden facial recognition surveillance embedded into everyday devices.
