Meta Adds Face-Recognition Code to Smart-Glasses App

WIRED published a technical review finding unreleased face-recognition code, internally named NameTag, embedded in the Meta AI app, the companion app used with Ray-Ban and Oakley smart glasses. WIRED reports the app has been downloaded more than 50 million times and that core components were added to the live app as early as January. Per WIRED, the functionality would detect faces, crop them, convert them into biometric faceprints, compare them against faceprints stored on a user's phone, and surface a notification when a match occurs. Cooper Quintin, a security researcher at the Electronic Frontier Foundation's Threat Lab who reviewed the code for WIRED, described the implementation as "nearly ready to go." Researchers who examined the code emphasize that no part of NameTag is currently running or transmitting biometric data to Meta's servers; it is dormant code, not a live feature.

Reporting describes three on-device model roles present in shipped builds: face detection, face cropping or segmentation, and encoding faces into biometric signatures. Keeping inference local reduces some server-exposure risk but preserves a significant biometric attack surface on user devices, and WIRED notes the phone-side database is configured to accept updates from Meta.

The code's presence is consequential because of Meta's legal history. WIRED notes Meta sunset its large-scale faceprint dataset in 2021 and paid a $650 million class-action settlement, and reports a $1.4 billion settlement the company agreed to in 2024 in related litigation. Civil-society opposition is vocal: more than 70 groups including the ACLU have publicly urged Meta to halt or disavow the feature, and EPIC has asked the FTC and states to block it.

Key signals include whether Meta enables the feature in production, how the phone-side faceprint database is provisioned and updated, whether the company publishes opt-in controls, retention windows, and a technical privacy specification, and any action from state attorneys general or the FTC. For practitioners, the episode reinforces that embedding dormant biometric capabilities into consumer apps can trigger legal and reputational scrutiny even before a feature ships, so similar work warrants high-scrutiny threat models, clear consent flows, and auditable local data handling.