A face-recognition system that Meta says does not exist has, for months, been sitting quietly inside an app on tens of millions of phones. The confusion at the center of this NameTag saga comes down to how Meta's own executives define two ordinary words, feature and exist. One senior figure at the company argues the software effectively is not real because ordinary users cannot switch it on, yet the same company's chief technology officer spent several minutes on a podcast last week describing exactly how the tool would work.
Code that sat inside a popular app for months
On June 4, it came to light that Meta had quietly built robust, though inactive, code for NameTag directly into Meta AI, the companion app for its Ray-Ban smart glasses that has been downloaded tens of millions of times. Andy Stone, Meta's vice president of communications, pushed back on X, arguing that several questions about how the feature would work had gone unanswered simply because the feature doesn't exist. Barely a day after that post, on June 5, Meta stripped the NameTag code out of the Meta AI app entirely.
Examination of the Meta AI app showed that pieces of NameTag's code had already appeared as early as January. By mid-February, The New York Times had reported separately that Meta was working on a NameTag face-recognition system, well before June's disclosure. By May, the core building blocks of that same code were fully present inside the app. Whether something so thoroughly built out counts as an existing feature depends entirely on how loosely or strictly Meta wants to define those two words on any given day.
A working demo: a system that actually recognized a face
To settle the argument in practical terms, a researcher who goes by the name Buchodi reviewed the code and was able to get the NameTag system to correctly recognize a photograph of philosopher Michel Foucault, a thinker famed for his writing on surveillance as a tool of power. That a stripped-out chunk of code could still positively identify a face is difficult to square with the claim that no functioning feature ever existed.
Bosworth spelled it out in detail on a podcast
The idea that Meta had no way of even describing how NameTag might work collapsed further just a week later. On the July 8 episode of the podcast The Most Interesting Thing in AI, host Nicholas Thompson, CEO of The Atlantic, put a segment of the conversation squarely under the heading What's true and false about NameTag and asked Meta CTO Andrew Boz Bosworth who exactly the system would identify. Bosworth answered at length: “Somebody you met in person with your glasses on who introduced themselves, or you said, OK, this is David, remember this person. Only available to you when you're wearing your glasses, this is a person you've met before. Here's their name. They're right in front of you... That's what we call a NameTags feature.” Later in the same conversation, Bosworth added, “So, it's a thing that, um, I think would be a great feature.”
Meta leans on a single word, would
Faced with the apparent contradiction between Stone insisting the feature doesn't exist and Bosworth describing it for minutes at a stretch, Meta has repeatedly pointed to the conditional grammar Bosworth used. Spokesperson Ryan Daniels went as far as bolding and underlining the word would while addressing the discrepancy. “There is no contradiction. Boz says this would be a good feature, particularly to answer the blind and low-vision community members' calls to help them identify people they've already met or want to remember,” Daniels said in a statement. “While we're exploring this, it's not available to consumers today. We think it's important that people understand this remains distinct from connecting glasses to a central database of people in the world, which is not a capability we are building.”
Here is what is not in dispute: NameTag existed, in a fully built, technically working form, as of roughly six weeks ago. Meta had been developing it since early 2025, licensing face-recognition software from a third party, assembling a complete detection-and-matching pipeline, and quietly loading it onto the tens of millions of phones running the Meta AI app, where it sat undetected until it was disclosed. Ordinary users could not switch it on without specialized tools, but analysis of the Meta AI code, backed up independently by two outside experts, confirmed a technically functional face-recognition system was living inside an app already installed on millions of devices. Taken at face value, that is the very system Bosworth spent minutes describing on the podcast.
Meta calls the coverage dishonest
Even so, Meta maintains, without saying exactly why, that describing NameTag as something that already existed is not accurate. The June 4 disclosure explicitly called the NameTag system unreleased in its opening lines and repeated that description throughout, yet Stone argued on X that this detail had been left unclear, calling the coverage “more than shoddy reporting” and “intellectually dishonest,” adding that it amounted to “pure advocacy-driven click bait.” Bosworth backed him up, describing the same coverage as “incredibly misleading” and “absolutely dishonest.” Meta has not explained what, specifically, its executives considered misleading or dishonest.
Why central database is the phrase Meta keeps repeating
One point Meta clearly cares about is the idea of a central database. During his conversation with Thompson, Bosworth insisted NameTag will not rely on any central database, even though nobody had actually made that claim. Analysis of the app instead found that the NameTag system converts faces captured through Meta's glasses into unique numerical signatures known as faceprints. Those faceprints could then be checked against a face database stored locally on a user's own device, one that Meta's servers had populated in the first place.
That distinction between one central database and millions of individual local databases, each sitting on a different phone but still connected back to Meta's servers, is not just a technicality. Bosworth himself pointed out that state laws such as Illinois's Biometric Information Privacy Act, or BIPA, and Texas's Capture or Use of Biometric Identifier Act, or CUBI, restrict how companies can publicly roll out face-recognition technology, generally requiring explicit consent before anyone's face is captured by such a system.
A feature Meta has been burned by before
Bosworth has good reason to tread carefully here. Back in 2019, Meta shut down its automatic face-recognition feature on Facebook, known as Tag Suggestions, after agreeing to a five billion dollar settlement with the Federal Trade Commission over privacy issues that included face recognition, and shortly before a separate 650 million dollar settlement with the state of Illinois specifically over Tag Suggestions.
By keeping the actual face-matching process on a user's own phone instead of querying a central server, Meta may be trying to position NameTag so it can argue the design complies with the letter of biometric privacy laws like BIPA and CUBI, if and when it ever ships the feature. Whether that kind of on-device design genuinely holds up in court, though, is far from settled.
Courts have split on the on-device defense
In 2021, a federal judge allowed a BIPA class-action lawsuit targeting Apple's Photos app to move forward, ruling that a company could plausibly be said to possess faceprints even when those faceprints are stored on a user's own device, since legal possession does not require exclusive control. That case was certified as a class action in June 2026 and is currently awaiting rulings on several pending motions.
Other courts have gone the opposite way. An Illinois appellate panel ruled in 2022 that Apple did not possess Face ID data that was stored solely on users' own devices. A federal judge separately dismissed a nearly identical lawsuit over Samsung's photo app in 2024, on the grounds that Samsung itself never received or accessed the face data that its own software had generated. Across these rulings, the real dividing line has not been where the data physically sits, but who actually controls it, whether using the feature is optional, whether a user can turn it off, and whether the company itself is ever capable of reaching the data at all.
The questions Meta still won't answer
When asked in June whether NameTag would be opt-in, and how the system stores faceprints and cropped images once they are captured, Meta declined to answer. On the single question courts have treated as the most decisive one, whether the data ever leaves a user's device or can be reached by anyone besides that user, Meta would not explain why it had licensed a third party's face-recognition software in the first place, when that licensing arrangement began, or whether it is still in effect today. Meta has also not addressed, even this week, why it keeps insisting that its face-recognition technology does not rely on a central database, when no one had suggested otherwise to begin with.
So does NameTag exist
Strip away the wordplay around feature and exist, and the underlying facts are straightforward. Meta designed and built a working face-recognition system, placed it, at least for a while, on the phones of millions of people, and appears to be seriously weighing whether to switch it on for good, all while one of its most senior executives spent several minutes publicly praising exactly what it would do. Whether that adds up to NameTag existing is, at this point, left for each reader to decide.
