Meta's Smart Glasses Hide Dormant Face ID Code—Again

Imagine walking into a room and your glasses whispering names at you. Creepy? Convenient? Both. That image is exactly why a fresh code discovery in Meta's smart-glasses companion app has privacy watchers sharpening their pencils.

Researchers say the app contains dormant facial-recognition code labeled 'NameTag' that has been quietly present in multiple updates since January 2026. This isn't some optional utility tucked away in a niche download. The app is required for Ray-Ban Meta models and related eyewear, and it has seen more than 50 million installs—meaning dormant code sits on a lot of phones.

What did the investigators find? The buried logic could use the glasses' camera to detect faces, create biometric templates, and compare those templates against a database stored locally on a user's phone. If a match appears, the wearer could receive an on-screen notification identifying the person. Short-circuiting privacy expectations, this design avoids a cloud lookup and instead keeps comparisons on-device—but local storage still carries risk.

Context matters. Meta publicly dismantled Facebook's face-recognition system in 2021 and said it deleted over a billion faceprints after backlash, regulatory pressure, and lawsuits. The company paid a roughly $650 million settlement in Illinois and later a $1.4 billion settlement stemming from biometric privacy claims in Texas. Those scars explain why any sign of a revival draws quick attention.

The code snippets reportedly point to three separate AI models tied to face detection and processing, plus traces of a user-facing feature meant to help people remember who they've met—an oddly human ambition for a machine. One internal test even generated a mock recognition alert using a face template attributed to the philosopher Michel Foucault, a detail that underlines how granular the trials were.

Meta says the traces reflect internal experimentation, not a product launch, and that it isn't building a centralized facial-recognition database. Company spokespeople stress no final decision has been made and promise transparency and caution if the idea ever moves forward.

Still, the discovery will likely reignite a debate that never really went away: do we prize convenience over the risk of constant identification? Wearable AI promises helpful nudges. But it can also blur the line between assistance and surveillance. Where that line falls is a policy question—and a personal one—worth watching closely.