Overview
- A WIRED code analysis published Thursday showed unreleased face‑recognition components called “NameTag” were embedded in the Meta AI companion app that pairs with Ray‑Ban and Oakley smart glasses.
- Security researchers including the EFF reproduced key behavior and found three models — for face detection, cropping, and encoding into 2,048‑number faceprints — that Meta had pushed from its servers to users’ phones.
- Tests by independent researchers reportedly produced a “Person recognized” notification after a faceprint was manually added to the app, indicating the system is close to functional even though the company has not enabled the feature.
- Meta told reporters it has not shipped an active feature and says it is not building a central face database, but it declined to answer which faces would be enrolled, how profiles would be created, or whether data might ever leave devices.
- Privacy groups, former Meta insiders and lawmakers warn the buried code revives past biometric litigation and could enable covert identification by abusers or other actors, and regulators and advocates are calling for clear limits or abandonment.