Meta moved fast. One day after reporters uncovered dormant code for a facial recognition system tucked inside its Meta AI companion app, the company pushed out an update that erased every trace of it. The discovery, detailed by WIRED on June 4, 2026, revealed that software capable of turning faces captured by Ray-Ban Meta smart glasses into biometric identifiers had been distributed to millions of phones since January.
By June 5 the components were gone. No explanation accompanied the removal. And no promise that the project known internally as NameTag was dead.
The episode highlights tensions that have followed Meta for years. After shuttering its Facebook photo-tagging facial recognition system in 2021 and deleting more than a billion faceprints, the company now finds itself circling back to the same technology. This time the hardware sits on people’s faces. The cameras point outward. The potential for silent identification in public spaces feels immediate.
According to the WIRED analysis, the code contained three separate AI models. One converted a photo of a face into a biometric template stored on the device. Another matched new camera feeds against that stored database. A third would trigger an alert to the wearer when a match occurred, displaying the recognized person’s name directly in the glasses. All processing happened locally. No cloud upload required for the core function. Still, the system needed an opt-in database of faces. Whose faces? The code left that question open.
Meta had signaled caution earlier. In April more than 70 advocacy organizations, including the American Civil Liberties Union, the Electronic Privacy Information Center and Fight for the Future, sent a letter demanding the company abandon NameTag. They warned it would hand powerful new tools to stalkers, abusive partners and anyone seeking to identify strangers without consent. The ACLU-led coalition called the feature a red line society must not cross.
At the time a Meta spokesperson told WIRED the company did not offer such a product. Any future version would receive a thoughtful approach. The discovery of already-shipped code undercut that message. It showed development had advanced far enough to reach production builds of an app downloaded more than 50 million times.
Back in February The New York Times reported that Meta aimed to bring facial recognition to its smart glasses partnership with EssilorLuxottica, owner of Ray-Ban and Oakley, as soon as this year. Four people familiar with the plans described an internal memo suggesting political distractions in the United States might mute criticism. The feature would let wearers identify people and pull information through the AI assistant.
That report carried echoes of earlier hesitation. Meta had considered facial recognition for the first generation of Ray-Ban Meta glasses in 2021 but backed away over technical and societal concerns. The 2021 shutdown of its broader facial recognition on Facebook came after years of backlash, a $650 million Illinois class-action settlement and a separate $1.4 billion agreement with Texas over biometric data collection.
Yet the hardware keeps advancing. Current Ray-Ban Meta glasses already record video and audio with a subtle LED indicator. They stream to the phone app. They integrate Meta’s AI for real-time queries. Adding name recall seems a small step technically. The societal distance feels vast.
Privacy advocates reacted quickly to the WIRED story. They pointed out that on-device storage does not eliminate risk. A lost phone could expose an entire social graph of recognized faces. Wearers could build databases of colleagues, protesters, romantic interests or political opponents without their knowledge. And the always-available camera changes social norms. People might self-censor simply knowing identification is possible.
Recent coverage reinforces the pressure. Engadget noted the rapid removal and tied the code directly to the NameTag project first surfaced by The New York Times. PetaPixel reported the same sequence, underscoring how quickly Meta acted once the code became public. Discussions on X, formerly Twitter, mixed alarm with skepticism. Some users called it inevitable surveillance creep. Others questioned whether the removal was genuine or merely cosmetic.
Meta’s communications chief Andy Stone described the work as purely exploratory. No final decision had been made on whether to proceed. The company has not commented further since the code purge. That silence leaves open the possibility of future attempts, perhaps with different architecture or after additional privacy safeguards.
The broader market context adds weight. Smart glasses sales are climbing. Meta and its partners have sold millions of the current models. Analysts expect the category to expand as displays improve and AI capabilities grow. Yet consumer trust remains fragile. Reports of contract workers in Kenya reviewing sensitive footage captured by the glasses, including private moments, have already damaged confidence. Adding facial recognition could intensify regulatory scrutiny from the Federal Trade Commission, state attorneys general and European data protection authorities.
Industry watchers see this episode as instructive. Companies can embed ambitious features in widely distributed software before public debate. When caught, they can retract without admitting error. The pattern suggests facial recognition for consumer wearables sits at the edge of acceptability. Technical feasibility has outrun societal agreement.
Meta faces a choice. It can continue refining NameTag in private until political or market conditions seem favorable. Or it can decide the privacy costs outweigh the engagement benefits and walk away for good. The speed of the code removal hints at sensitivity to immediate backlash. But history shows the idea itself has proven hard to kill.
For now the biometric templates stay off millions of phones. The glasses continue as stylish cameras and AI companions without the ability to name strangers on sight. That limit may not last. Public pressure, lawsuits and legislation could force clearer boundaries. Or Meta could thread the needle with strict consent mechanisms, transparent databases and on-device guarantees that satisfy both users and critics.
Either path carries risk. The technology exists. The hardware is in circulation. And the discovery of hidden code has made one thing obvious. When it comes to facial recognition in everyday glasses, Meta isn’t starting from scratch. It has already taken several quiet steps down the road.
WebProNews is an iEntry Publication