Meta Pulls Facial Recognition Code From Smart Glasses App After Quiet Deployment

Meta built a facial recognition system for its Ray-Ban smart glasses. Then it shipped the code to millions of phones without telling anyone. Discovery forced a swift reversal.

The episode reveals ongoing tensions inside the company over privacy, ambition and public backlash. Executives once promised to walk away from facial recognition. Events this week show that retreat was temporary.

On June 4, WIRED reported that an unreleased feature called NameTag sat inside the Meta AI companion app, downloaded more than 50 million times. The app powers key functions for Meta’s smart glasses, including the popular Ray-Ban and Oakley models. Code analysis uncovered libraries explicitly tied to face recognition. The system would let wearers identify strangers in real time and pull personal details through the AI assistant.

One day later, Meta stripped those components out. A new version of the Meta AI app published Friday contained none of the face recognition libraries present in the previous build, according to further WIRED analysis of the code. The removal came without explanation or announcement from the company.

But the move came too late to avoid scrutiny. Privacy advocates had already sounded alarms months earlier when The New York Times first exposed Meta’s plans for the feature. In February, the paper reported that Meta aimed to roll out NameTag as soon as this year. Four people involved with the plans described an internal belief that the current political climate would mute opposition. Civil liberties groups saw it differently.

More than 75 organizations, led by the ACLU, sent a letter to Mark Zuckerberg in April. They called the proposed technology a red line society must not cross. The coalition warned that NameTag would hand stalkers, abusers and law enforcement the power to silently identify anyone in public. Sexual predators could target victims. Agents could scan crowds without warrants. The backlash was swift and broad.

Meta’s history with facial recognition adds weight to the concerns. In 2021 the company announced it would shut down its photo-tagging system on Facebook and delete more than a billion stored faceprints. Privacy scandals and regulatory pressure drove the decision. At the time, executives framed it as a permanent shift. Jerome, a source who joined Reality Labs in 2021, told WIRED there was always internal tension about when the technology would return.

“There was always this tension of, well, when do we roll back out face recognition?” the former employee said.

That question appears answered for now. Yet the code’s presence in a production app suggests development had advanced far beyond conceptual stages. NameTag was not some distant research project. It had reached the point of integration into software already running on user devices. And it revived capabilities Meta once claimed to abandon.

The Gizmodo report captured internal frustration at the reversal. Executives were reportedly mad about the need to pull the system. Public statements from the company remain measured. A Meta spokesperson told multiple outlets the company continues to think through options and would take a thoughtful approach before any rollout.

But actions speak louder. The quiet embedding of dormant code followed by its rapid deletion after exposure points to a pattern. Meta tests boundaries. When caught, it retreats. Then it waits for the next opportunity.

Smart glasses sales have soared despite privacy fears. Users love the cameras, speakers and AI features. Yet those same cameras make the devices ideal for surreptitious recording. Add real-time facial recognition and the glasses transform from stylish accessory to powerful surveillance tool. A wearer could glance at someone in a bar, coffee shop or protest and receive instant background information. Consent never enters the equation.

Earlier demonstrations highlighted the risks. In 2024 two Harvard students rigged Ray-Ban Meta glasses with commercial facial recognition services like PimEyes. They identified strangers on the subway. Videos went viral. Meta insisted at the time that its hardware did not perform recognition. The students relied on third-party cloud services. That distinction offered little comfort to those filmed.

Now the company had moved toward building the capability itself. The NameTag system would operate through Meta’s own AI, tying identification directly to its vast social graph and data stores. The privacy implications multiply.

Advocates argue this crosses a fundamental line. Public anonymity erodes when anyone with a pair of glasses can unmask you. Women face heightened risks from stalkers. Protesters lose plausible deniability. Everyday interactions change when faces become searchable data points.

Meta’s glasses already sparked incidents. A Border Patrol agent wore them to an immigration raid. A university warned about a man filming female students. An esthetician was caught recording clients during intimate procedures. These cases involved basic camera use. Facial recognition would amplify the harm.

Regulatory pressure continues to build. The Electronic Privacy Information Center urged the FTC and state attorneys general to block the feature before launch. Letters cited existing privacy risks from the glasses’ always-on capabilities. Real-time identification would only compound them.

So Meta removed the code. For now. The company has not ruled out future deployment. Its history suggests persistence. After the 2021 shutdown, internal discussions never treated the exit as final. Development continued in some form, culminating in the libraries discovered this month.

The speed of the reversal after the WIRED report raises questions about readiness. Was the feature close to activation? Or did Meta simply want the option in place? Company silence leaves observers to speculate. What seems clear is that pressure from journalists and advocates still shapes outcomes.

Industry watchers note the broader context. Wearables represent the next frontier after smartphones. Meta, Apple and others race to define it. Features that feel futuristic today become table stakes tomorrow. Yet biometric identification in public spaces carries unique dangers that phones never did.

Consumers have proven willing to trade some privacy for convenience. Smart glasses sales demonstrate demand. But the backlash to NameTag shows limits exist. Not every innovation finds acceptance.

Meta faces a choice. It can pursue aggressive deployment and fight regulators, advocates and users. Or it can accept that some capabilities, however technically feasible, cross societal lines. The recent removal suggests the latter view gained ground this week. Whether it holds remains uncertain.

One thing is clear. The technology exists. The code was written, tested and distributed. Deletion from one app version changes little about underlying capabilities or corporate intent. Privacy advocates vow continued vigilance. Meta’s next move will face immediate examination.