What Meta’s Hidden NameTag System Was Designed to Do
Meta smart glasses facial recognition refers to the dormant NameTag feature whose code inside Meta’s AI companion app could detect faces, generate biometric faceprints, and identify people through data stored directly on a user’s phone without clear notice or consent. Reverse engineering of the Meta AI app, internally called Stella, revealed three machine-learning models working together to detect, align, and embed faces into 2,048‑dimensional biometric fingerprints. This facial recognition code went far beyond camera autofocus, tying into an SQLite vector database for similarity search, persistent storage for unknown faces, and a notification system that could announce when a person was recognized. Researchers demonstrated an end‑to‑end match by identifying a portrait of philosopher Michel Foucault after loading his biometric data, showing the system was technically functional even though Meta had not exposed a user-facing NameTag feature.
How WIRED and Researchers Exposed Faceprint Tracking
The controversy ignited when WIRED reported that Meta had embedded substantial portions of NameTag in the Meta AI app used by tens of millions of Ray-Ban Meta smart glasses owners. Security researchers at Buchodi and analysts at Gadget Review documented a complete pipeline for faceprint tracking: faces captured by the glasses, processed on-device, and stored as cropped images plus binary embeddings in a private directory that survived reboots. The app also contained hidden strings and screens for a “Connections” feature that would “remember the people you met,” as well as deep links to profile pages that were not visible to normal users. Privacy advocates and the Electronic Frontier Foundation’s Threat Lab later confirmed the technical findings, arguing that shipping such infrastructure without prominent disclosure undermines trust in Meta’s approach to privacy surveillance.
Meta’s 24-Hour Cleanup and Confusing Messaging
Within about 24 hours of WIRED’s exposé, Meta pushed an update to the Meta AI app that removed key parts of the facial recognition code. According to Gadget Review, “Meta removed face recognition libraries, alert systems, and storage locations for biometric data in the updated app version,” leaving only scattered debug fragments and dead menu entries pointing to missing components. Publicly, Meta has described NameTag as internal experimentation and claimed that “nothing has shipped to consumers,” while simultaneously criticizing the reporting as dishonest. The company says it is not building a centralized facial recognition database and has promised to move carefully with any such feature. Yet Meta has not given direct answers about what test data was collected, where it may have been stored, or whether it was wiped along with the code.
Why Privacy Advocates See a Surveillance Warning Sign
For privacy advocates, NameTag is less a one-off experiment than a warning about how easily smart glasses can become personal surveillance systems. The dormant feature would have turned Meta smart glasses into always-on tagging tools, quietly building a private gallery of faces—recognized and unknown—linked to biometric fingerprints on each device. Even if processing remains local, Meta’s own disclosures say AI glasses data may be stored on both devices and remote servers, raising concerns about future data flows. The Electronic Frontier Foundation’s Threat Lab and other researchers argue that embedding a nearly launch-ready recognition pipeline without clear consent shows why strong biometric rules are needed. Given Meta’s history of Facebook face recognition shutdowns and legal settlements over faceprints, the discovery reinforces doubts about whether the company’s current promises are enough to protect people from pervasive privacy surveillance.
Could Meta’s NameTag Feature Return in Another Form?
Meta’s decision to strip NameTag code after exposure does not settle whether facial recognition will return to its smart glasses ecosystem. The company has said no final decision has been made and that the current code reflects experimentation, but the level of polish—multiple models, databases, notifications, and hidden UI—suggests it was close to being product-ready. Future reintroductions might emphasize on-device processing, opt-in prompts, or limited contact-based recognition to ease concerns. Yet critics note that consent is hard to guarantee when bystanders never agreed to have their faces scanned or stored. For now, Meta smart glasses ship without user-accessible facial recognition, but the technical groundwork and business incentives remain. The real test will be whether Meta designs any future NameTag-style feature with clear disclosure, granular control, and strict limits on faceprint tracking.
