What Meta’s Hidden Facial Recognition Feature Was Meant To Do
Meta facial recognition in the Meta AI smart glasses app refers to dormant code that could scan faces through wearable cameras, convert them into biometric data stored on a user’s phone, and push alerts identifying people nearby, all without a clear opt‑in from those being scanned. The internal feature, known as NameTag, relied on three AI models: one to detect faces, one to crop them, and one to encode them into biometric identifiers ready for comparison. Meta said the work was experimental and that nothing had shipped to consumers, yet the code shipped inside an app required for Ray‑Ban Meta and other smart glasses that has been installed tens of millions of times. For privacy advocates, that combination of biometric data tracking and inconspicuous deployment turned an experiment into a live smart glasses privacy test case.
Timeline: From Quiet Deployment to WIRED’s Discovery and Backlash
According to reporting cited by multiple outlets, facial recognition code tied to NameTag appeared in several Meta AI app updates starting in January 2026, years after Meta had announced the shutdown of Facebook’s earlier face-recognition system. Researchers who examined the app found it was nearly ready to identify people through smart glasses, with on-device models for detection, cropping, and encoding faces plus traces of a user-facing recognition alert. One test even used a face template based on philosopher Michel Foucault to generate a sample notification. The feature remained disabled, but the presence of live code in an app downloaded more than 50 million times gave critics grounds to argue that Meta had crossed a line between internal testing and quiet rollout. Civil society groups, already warning about surveillance technology concerns in wearables, quickly amplified the findings.
Meta’s Response and the Sudden Facial Recognition Removal
After WIRED reported on the dormant facial recognition system, Meta responded that it was “exploring” the technology and had not decided whether to release NameTag. The company also insisted it was not building a central face database, framing its work as on-device experimentation tied to smart glasses privacy features. But within roughly 48 hours of public outcry and scrutiny from researchers and digital rights groups, Meta pushed a June 5 update to the Meta AI app that removed the face-recognition models, “Person recognized” alert logic, and related biometric processing code. The Electronic Frontier Foundation’s Threat Lab confirmed through static analysis that the facial recognition removal was extensive, wiping out the pipelines that converted faces into stored identifiers. Despite that rollback, Meta declined to say whether NameTag might return, or what happened to any data created during internal testing.
What This Reveals About Testing, Oversight, and Smart Glasses Privacy
The NameTag episode highlights a structural problem: sensitive surveillance technology concerns are often handled through opaque “experiments” that blend into consumer apps long before public debate. Meta embedded code capable of biometric data tracking into devices people wear in public spaces, without a clear way for bystanders to consent or opt out, and only reversed course once the system was exposed. Critics argue that relying on internal ethics reviews and marketing promises of a “thoughtful approach” is not enough when smart glasses can silently identify strangers. Past legal battles over biometric data, including large settlements around earlier Facebook face-recognition practices, show that after-the-fact penalties do not prevent new attempts. The swift deletion of the NameTag system under pressure suggests that external oversight, enforceable privacy rules, and transparency about on-device experiments are becoming essential safeguards.
The Future of Wearable AI and Biometric Data Tracking
Even without an active NameTag feature, Meta facial recognition research for wearables exposes a direction of travel: smart glasses are edging toward constant recognition, not just recording. Supporters say on-device identification could help people remember names or assist users with memory or vision impairments, but privacy advocates warn of a de facto distributed surveillance network if similar tools spread without strict limits. Meta says it is not building a central face database, yet on-phone collections of biometric signatures can still be misused or repurposed. For now, the facial recognition removal shows that public pressure can still slow ambitious biometric features, but it does not resolve how or when tech companies should test them. Until clear rules govern smart glasses privacy and biometric data tracking, each new experiment risks turning users and bystanders into subjects in unannounced trials.
