In early June 2026, a bombshell investigation by WIRED revealed that Meta had quietly embedded an unreleased facial recognition system inside its Meta AI app — the very app used by owners of Ray-Ban and Oakley Meta smart glasses. The code, internally referred to as “NameTag,” was never announced. There was no press release, no opt-in prompt, and no public explanation. Yet the underlying technology had been silently distributed to the phones of over 50 million users across multiple app updates, dating as far back as January 2026.
This story isn’t just about a tech company testing a new feature. It’s about what happens when a company with Meta’s history around privacy pushes powerful biometric surveillance technology into everyday wearables — quietly, deliberately, and without your knowledge.
What Is NameTag, and How Does It Work?
At its core, NameTag is a face-recognition pipeline engineered to run entirely on your smartphone. WIRED’s investigation, backed by independent security researchers who reviewed the companion app’s code, found that Meta had embedded three distinct AI models within the app:
- Model 1 detects faces captured by the glasses’ cameras
- Model 2 crops and isolates those faces
- Model 3 converts them into biometric “faceprints” — unique mathematical representations of a person’s face
These faceprints are then stored locally on the device and compared against a local index. If there’s a match, the system would send the glasses wearer a notification — essentially telling them who they’re looking at in real time.
A security researcher who analyzed the app confirmed they were able to run this pipeline end-to-end on a test image: the system detected a face, generated a 2,048-dimension biometric embedding, searched a local index, and fired an Android notification reading “Person Recognized.”
By May 2026, the feature had been rebranded internally as “Connections,” with user-facing interface text inviting people to “remember the people you met.” Two potential versions of the feature reportedly exist: one that recognizes only existing contacts on Meta’s platforms, and a more expansive version that could identify anyone with a public profile on Instagram or Facebook.
As of now, the feature is not active for consumers. But the machinery is fully built, wired, and sitting dormant on tens of millions of phones.
Why the Secrecy? The Leaked Memo Tells a Disturbing Story
What makes this story especially troubling is not just the technology itself — it’s the internal reasoning behind how Meta planned to launch it.
Back in February 2026, The New York Times reported on a leaked internal Meta document that revealed the company was eyeing a year-end rollout of NameTag. More alarming was the strategic framing within that memo: Meta reportedly saw the current “dynamic political environment” as an opportunity, reasoning that civil society groups who would normally challenge such a launch would have their attention — and resources — focused elsewhere.
In plain terms: Meta was banking on critics being too distracted to push back.
Privacy advocates were quick to sound the alarm. The Electronic Frontier Foundation (EFF) and the Electronic Privacy Information Center (EPIC) both urged federal and state regulators to investigate before any launch. In April 2026, the ACLU led a coalition of 75 organizations in an open letter to CEO Mark Zuckerberg, calling the feature an “unacceptable threat to privacy and liberty” and demanding the project be scrapped entirely — not modified, not delayed, but cancelled outright.
The On-Device Strategy: A Legal Workaround?
One of the most technically interesting — and legally strategic — aspects of NameTag is where the data lives. Unlike traditional facial recognition systems that upload images to remote servers, Meta’s system stores biometric faceprints locally on the user’s phone.
Privacy researchers believe this is not a coincidence. Illinois’ Biometric Information Privacy Act (BIPA), one of the strongest biometric privacy laws in the United States, primarily regulates how companies collect and store facial data on their own servers. A system where data remains on a user’s device technically occupies a legal grey zone under that law — the same law that has cost Meta $650 million in a class-action settlement in 2021 and contributed to a broader $7 billion total in privacy-related legal costs for the company.
By moving the biometric infrastructure onto your device, Meta may be engineering a system that sidesteps existing legal guardrails while still delivering the same real-world capability: identifying strangers in real time.
Meta’s Complicated History With Face Recognition
To understand the weight of this moment, it helps to remember where Meta has been before.
- In 2021, Meta shut down Facebook’s photo-tagging facial recognition system, deleting the faceprint data of over one billion users. The company cited “ongoing uncertainty” in global regulation as the reason.
- In 2019, Meta (then Facebook) agreed to a $5 billion FTC settlement that included allegations around deceptive facial recognition settings.
- In 2024, facial recognition quietly returned to Instagram and Facebook — this time framed as a tool for detecting scam ads that misused celebrities’ likenesses.
- In 2026, Meta is now actively building the feature into always-on, eye-level wearable cameras.
The pattern is clear: Meta’s relationship with facial recognition has never truly ended. It has simply gone quiet between cycles — and each time it returns, the technology is more capable and more embedded in daily life.
The Bystander Problem: You Don’t Have to Own the Glasses to Be Affected
Perhaps the most overlooked dimension of this entire issue is who bears the real risk — and it’s not necessarily the people wearing the glasses.
Consider: more than 7 million pairs of AI smart glasses were sold globally in 2025 alone, with Meta holding over 80% of that market. If researchers’ projections hold, as many as 100 million people could own smart glasses in the coming years. That means the bystander problem — the risk to ordinary people who simply exist in public spaces near someone wearing these glasses — is not theoretical. It’s already arriving.
Women have reportedly been filmed at beaches and outside shops by men wearing the glasses, sometimes while appearing to hold a normal conversation. Retail workers have been recorded during staged pranks. Drive-through staff have been captured without their knowledge. None of these cases required face identification. Add NameTag to the equation, and every one of these strangers becomes identifiable by name, social profile, and workplace — without ever consenting to be scanned.
Meta spokesperson Tracy Clayton addressed the bystander issue by stating that “the onus is ultimately on individual people to not actively exploit” the technology — a response that privacy advocates widely criticized as inadequate given the scale of potential misuse.
Government Use Adds Another Layer of Concern
The stakes were raised further when it emerged that ICE agents in at least six U.S. states had been spotted wearing Meta’s Ray-Ban smart glasses during immigration enforcement operations in early 2026. While NameTag is not active in those glasses today, the Department of Homeland Security has separately requested $7.5 million to develop its own smart glasses with real-time facial recognition capability for field agents — with a target delivery date of September 2027.
The convergence of commercial wearable tech and government surveillance infrastructure is no longer a hypothetical. It is an active development, and Meta’s NameTag sits at the center of that intersection.
What Meta Says — and Doesn’t Say
Meta has acknowledged that it is “exploring” facial recognition for its smart glasses but maintains the feature is not active. The company has not provided a timeline for launch, details on what user controls would look like, or whether explicit consent would be sought before activating biometric identification through the glasses.
What makes this silence notable is that Meta’s previous statements publicly characterized the technology as something still being “thought through” — even as internal app updates were distributing the code to users’ devices month after month. The gap between what the company said publicly and what it was doing technically is significant.
Why This Matters for Everyone
The broader implications of what Meta is building extend far beyond any one company or product. Smart glasses — often visually indistinguishable from regular eyewear — are designed to be worn all day, in every environment. A single wearer could passively scan thousands of faces in a single afternoon. If NameTag activates, none of those people would know they had been identified.
That kind of ambient, frictionless identification does something new to public life. It eliminates the practical anonymity that ordinary people have always had when walking down a street, entering a café, or attending a public gathering. It transforms every social space into a potentially surveilled one, controlled not by governments or security agencies, but by any individual with a pair of glasses and an app.
The ACLU put it simply: “The American people have not consented to this massive invasion of privacy.”
The Bottom Line
Meta’s NameTag feature may not be live. But the code is on your phone. The AI models are trained and embedded. The notification infrastructure is wired and ready. And the company has, by its own internal reasoning, been waiting for the right moment — politically and competitively — to flip the switch.
Whether that moment comes this year, next year, or after the next regulatory cycle, one thing is clear: the question is no longer if real-time facial recognition comes to mainstream wearables. The question is when — and whether anyone will have agreed to it.