Meta Put Pentagon Face-Scanning Tech in 50 Million Phones — Then Quietly Deleted It

Imagine you're standing in line at a coffee shop. The person next to you is wearing a pair of trendy glasses — the kind that look totally normal. What you don't know is that those glasses may have just tried to figure out who you are. Not to record you. Not to photograph you. To identify you. By your face. Without asking.

That's not a sci-fi script. According to reporting cited by Digital Trends, that is exactly the direction one of the world's largest social media companies was heading — quietly, and until recently, without you knowing a thing about it.

This Wasn't a Research Project in a Lab Somewhere

Here's the part that should make you sit up straighter. Meta didn't just explore the idea of face recognition in glasses. According to the investigation covered by Digital Trends and Malwarebytes, the company actually embedded working face-recognition code — sometimes called "NameTag" code — directly into its companion app. That app sits on more than 50 million phones. The code was already there. On devices people actually own. In pockets and purses right now.

It was deleted only after someone noticed it was there.

Think about that sequence. Build it. Ship it quietly. Delete it when discovered. That's not how you handle a safety feature. That's how you handle something you weren't sure you wanted people to see yet.

And the contractor Meta brought in to help build this? Not a startup. Not a Silicon Valley computer-vision lab. According to PhoneArena, the supplier earns roughly 80% of its revenue from government clients — including the U.S. military and law enforcement agencies. This is defense-grade identification infrastructure. The kind built to find people in conflict zones or track suspects across surveillance feeds. And it was being prototyped for glasses that come in tortoiseshell and gold. This article is part of a series — start with One Stolen Badge Shouldnt Unlock Your Whole Office Heres Wha.

The Timing Memo Is the Part That Should Bother You Most

Look, companies prototype things and walk them back all the time. That's normal. What's not normal is what reportedly appeared in internal Meta memos from May 2025. According to the reporting, those internal documents suggested that facial recognition features should be added during a period when civil liberties groups would be "too busy to protest."

Read that again slowly.

That's not an engineering decision. That's a strategy. Someone inside the company looked at the calendar, identified a window of reduced public attention, and suggested using it to slip a face-identification feature into consumer hardware before anyone could organize a response. When a company starts scheduling its product rollouts around protest calendars, you're not dealing with a company that's uncertain about whether the feature is appropriate. You're dealing with a company that already knows the answer — and is hoping you don't get to weigh in.

"How does Meta plan to obtain the affirmative express consent of not just users of the device, but every person whose biometric data — the measurements and calculations uniquely linked to their face — may be captured in the glasses' field of view?" — Senators Markey, Wyden, and Merkley, U.S. Senate Official Press Release, March 2026

That question from three U.S. senators — posed in a formal letter demanding answers from Meta in March 2026 — is the one nobody has a good answer for. And that's because there isn't a good answer. Not yet. Maybe not ever, with the technology as it currently works.

The Consent Problem Nobody Wants to Say Out Loud

Here's what makes this different from a security camera at a gas station. Fixed cameras are attached to walls. You walk past one, maybe it records you, maybe someone reviews the footage later for a specific reason. That's uncomfortable, but it's bounded. There's a location. There's an operator. There's (usually) a sign.

Smart glasses are none of those things. They walk around with people. They go to weddings, school pickups, neighborhood barbecues. The person wearing them looks like anyone else. And if those glasses can run a face-identification check — comparing your face against a database to pull up your name, your social profiles, your employer — that check happens in seconds. Silently. Without any visible indicator. You will never know it occurred.

As Biometric Update's expert analysis put it: the real danger of wearable identification isn't one dramatic misuse. It's normalization. Once a capability exists in a device people carry everywhere, the threshold for using it keeps dropping — until it's just part of how people interact. Or how they're surveilled, depending on which side of the glasses you're standing on. Previously in this series: Your Kids Career Could Hinge On A Camera That Says Not You.

What Meta Says — And Why It's Not Quite Reassuring

To be fair: Meta's official position is that no face-recognition feature has shipped for the glasses. No final decision has been made. The company says it isn't building a central face database. The code, in their framing, was exploratory engineering — the kind of technical tinkering that happens constantly and never becomes a product.

That's possible. Software prototypes get abandoned all the time. (The graveyard of tech experiments is very large and very well-populated.)

But the pattern here doesn't feel like an abandoned experiment. It feels like capability-building with a very specific trajectory: license military-grade identification technology, embed it in a consumer app at scale, delete it when noticed, and wait for a better moment. Nothing about that sequence suggests the idea was abandoned. It suggests the idea is patient.

And the Senate inquiry pushed exactly that point — asking not just what Meta plans to do, but how on earth they would even make consent work for bystanders. People who walk past a glasses-wearer on a sidewalk. People at a party who didn't sign any terms of service. People who have no relationship with Meta whatsoever.

There is no checkbox for that. There is no popup notification. There is no version of "I agree" that covers a stranger's face being scanned while they wait for the crosswalk light to change.

What You Can Actually Do With This Information

If you've ever looked at a photo of someone online and wondered, "Is that really who they say they are?" — that question and this story are connected. The ability to match a face to an identity is genuinely useful when it's controlled, limited, and applied to specific cases where you're the one asking. Verifying a date's profile photo before meeting them. Confirming that a job applicant's headshot matches who showed up. That's a tool. That's a choice you made. Up next: Why Passkey Adoption Is Stalling Recovery Problem.

What's being built into glasses is something different. It's the same matching capability, unmoored from any specific question you asked — running in the background, on strangers, continuously, invisibly.

For now, the most useful thing you can do is stay literate about the gap between what facial comparison is designed to do (answer a specific question you asked, with your knowledge) and what ambient identification actually does (answer questions nobody asked you, about people who had no idea they were being checked). Those are not the same technology in different packaging. They are different things wearing the same face.

The senators asked Meta how consent would even work for bystanders. Meta has not answered that question in any public forum. Smart glasses shipments are growing at 210% a year. And the code — though currently deleted — was already sitting in 50 million pockets.

The next camera that tries to identify you may not look like a camera at all. It may just look like someone's glasses. And the person wearing them may not even fully understand what their frames are doing.

That's not a future problem. That's the current prototype.