What Meta’s Hidden Facial Recognition System Was Designed to Do
Meta facial recognition in its smart glasses ecosystem refers to the dormant NameTag feature and underlying facial recognition code secretly embedded in the Meta AI companion app, which was technically capable of detecting faces through Ray-Ban glasses, generating biometric identifiers, and storing them locally for later identification, all without meaningful disclosure or user-facing controls. Reverse engineering revealed a complete pipeline: three machine learning models detect faces, align them, and create 2,048‑dimensional biometric fingerprints tied to cropped face images on the phone. An SQLite vector database handled similarity matching, while a notification system was prepared to announce when a “person was recognized.” This hidden infrastructure turned the app into a potential Ray-Ban glasses surveillance tool, enabling biometric data tracking of both contacts and strangers under the label of an “unreleased” experiment that users could not see or configure.
Timeline: From Facebook Shutdown to Silent NameTag Rollout
Meta publicly closed Facebook’s large-scale facial recognition system in 2021 and said it would delete more than a billion stored faceprints after regulatory pressure and costly biometric lawsuits, including a $650 million settlement in Illinois and a later $1.4 billion settlement in Texas. Yet researchers later found that code for NameTag had been present in multiple Meta AI app updates, quietly shipping with smart glasses support to tens of millions of Ray-Ban Meta owners. Security analysis showed the codebase included three ExecuTorch models, internal strings for a “Connections” feature to remember people, and test notifications that identified a portrait of Michel Foucault after his biometric template was preloaded. Even though NameTag remained switched off in the interface, the long-running presence of near-launch facial recognition code pointed to sustained internal interest long after the public shutdown.
WIRED’s Exposure and Meta’s 24-Hour Code Purge
The situation changed when WIRED reported that Meta had quietly shipped substantial NameTag components to millions of devices through the Meta AI smart glasses companion app. Within 24 hours of that report, Meta pushed an update that stripped out key elements of the facial recognition code. According to Gadget Review, Meta removed core libraries, biometric processing workflows, and the private storage directory where cropped faces and binary embeddings had been stored, leaving only scattered debug fragments and broken menu references. The company said that people should not over-interpret the discovery and framed NameTag as an exploratory feature, but it did not explain why such a complete system had been distributed for so long or why the cleanup was so rapid. The timing suggested an internal reassessment triggered by public scrutiny rather than a routine engineering decision.
Smart Glasses Privacy and Biometric Data Tracking Risks
The NameTag design shows how smart glasses privacy problems deepen when wearable cameras are paired with hidden biometric data tracking. Technically, NameTag would have allowed wearers to build local face databases of “people they have met,” receive real-time recognition alerts through their Ray-Ban glasses, and keep unknown faces on the device “for future processing.” Even without a central server storing faceprints, this model turns everyday encounters into entries in a personal surveillance ledger, often without the knowledge or consent of people being recorded. For bystanders, there is no obvious visual signal that their face could be turned into a 2,048‑dimensional biometric fingerprint. For wearers, there were no clear user controls, opt-in flows, or transparency tools in the companion app, despite the presence of the full backend facial recognition code path.
What Meta’s Cleanup Means for the Future of Face Recognition
Meta says it is only exploring facial recognition for smart glasses, denies operating a central face database, and stresses that NameTag never launched to the public. However, the discovery that near-finished facial recognition code shipped in production for years, followed by a fast and quiet deletion after exposure, leaves major questions unanswered. Could similar code return in a revised form? Will Meta disclose any future biometric experiments in its Ray-Ban smart glasses ecosystem before pushing them to millions of users? More broadly, the NameTag episode shows how powerful surveillance functions can hide behind ordinary app updates, out of sight of both users and regulators. Unless clear rules demand disclosure, consent, and strict limits on on-device face databases, smart glasses will remain a fault line in the debate over everyday biometric surveillance.
