Analysts Find Dormant Facial-Recognition Code in Meta’s Smart-Glasses App

Independent reporting and technical analysis published Thursday found that Meta has built dormant facial-recognition capability into the companion app for its smart glasses, including code that can turn a face into a 2,048-number biometric template and compare it against faces that have been manually enrolled.

The finding, however, comes with an important limit. WIRED, the Electronic Frontier Foundation’s Threat Lab and an independent researcher all said the feature is not exposed to ordinary consumers. There is no public evidence from their analyses that Meta’s smart-glasses users are currently identifying strangers in normal use or that Meta has pushed identity databases to their accounts.

The evidence came from three separate reviews released June 4. WIRED reported that unreleased face-recognition code, internally called “NameTag,” is present in Meta’s smart-glasses companion app and that core components were integrated into software distributed to users as early as January 2026. The Electronic Frontier Foundation, a digital rights group, said its Threat Lab confirmed through static analysis — examining software code without running it — that facial-recognition code is in the app. Independent researcher Buchodi published a separate analysis of the Android version and found a complete but dormant on-device recognition pipeline.

According to Buchodi, EFF and WIRED, the system creates a 2,048-number faceprint, also known as a biometric embedding, and compares new faces against templates stored in a local database on the phone. Buchodi said that after manually adding a face to the app’s database using a phone connected in debug mode, the system later detected that face and triggered an Android notification reading “Person recognized” and “Recognized .” Buchodi also found that when the system does not find a match, it stores the face crop and embedding locally on the device in persistent storage.

Buchodi drew a sharp distinction between what the code can do and what has been shown in public. “This is not ‘Meta is secretly identifying the people you look at,’” the researcher wrote. “It is: the complete apparatus to do exactly that is sitting on the device, assembled and functional, gated by Meta.”

The discovery matters in part because of the app’s reach. WIRED reported that the Meta AI companion app, which is required for key smart-glasses features, has been downloaded more than 50 million times. That means code containing the dormant capability is already present on many users’ phones, even if the feature is not available to them.

Meta told WIRED that the company remains in an exploratory phase. “One decision we can be clear about — we are not building a central face database,” a spokesperson said. WIRED also quoted Meta as saying, “we’re exploring these types of features … what you’re seeing is merely evidence of that exploration.”

That response is likely to draw close attention because Meta has a long and expensive history with facial recognition. The company shut down Facebook’s face-recognition system in 2021 after years of criticism and legal pressure. That same year, Meta agreed to pay $650 million to settle an Illinois lawsuit under the state’s Biometric Information Privacy Act, a law that regulates the collection and use of biometric data. In July 2024, Texas announced a $1.4 billion settlement with Meta over biometric-data claims.

Lawmakers are already asking questions about smart glasses and biometrics. On March 17, 2026, Sens. Edward Markey of Massachusetts, Jeff Merkley of Oregon and Ron Wyden of Oregon sent Meta CEO Mark Zuckerberg a letter seeking details about Meta’s facial-recognition and biometric-data practices for smart glasses.

The result is a finding with immediate regulatory significance but a narrow factual scope: Independent analysis shows the capability exists inside Meta’s app and works in a test environment, but there is no verified evidence that Meta has rolled facial recognition out to ordinary consumers.