Meta's NameTag Face Recognition: Built, Deployed, and Denied

Meta's NameTag Face Recognition: Built, Deployed, and Denied

Meta has spent weeks insisting that NameTag, a face-recognition system built for its smart glasses, does not exist. Yet code for the system was embedded in an app downloaded by tens of millions of people, a researcher successfully used it to identify a face, and Meta's own CTO described its functionality in detail on a podcast. The gap between Meta's public statements and the documented reality has reignited debate over how tech companies talk about unreleased features—and over the privacy implications of bringing face recognition to consumer wearables.

Code Found in Millions of Phones

On June 4, WIRED reported that Meta had included robust but inactive code for NameTag in Meta AI, the companion app for Meta Ray-Ban glasses. The app has been downloaded tens of millions of times. According to WIRED's analysis, NameTag code first appeared in the app as early as January. By May, the core components of the system were present.

Meta's vice president of communications, Andy Stone, responded to the report on X, writing that the feature "doesn't exist." Meta removed the NameTag code from the Meta AI app the following day. Stone later called WIRED's reporting "shoddy," "intellectually dishonest," and "pure advocacy-driven click bait." Meta CTO Andrew "Boz" Bosworth echoed those sentiments, calling the report "incredibly misleading" and "absolutely dishonest." Meta did not answer WIRED's questions about what specifically its executives found misleading.

Despite Meta's denials, a researcher who goes by the name Buchodi reviewed the code at WIRED's request and successfully used the NameTag system to recognize a photograph of philosopher Michel Foucault, known for his writings on surveillance as an instrument of power. WIRED's analysis, confirmed by two independent experts, found a technically functional face-recognition system within the app.

Executive Describes Feature in Detail

The claim that Meta could not describe how NameTag works was further complicated on July 8, when Bosworth appeared on the podcast "The Most Interesting Thing in AI," hosted by Nicholas Thompson, CEO of The Atlantic and a former WIRED editor in chief. During a segment labeled "What's true and false about NameTag," Bosworth explained that the system would identify "somebody you met in person with your glasses on who introduced themselves—or you said, 'OK, this is David, remember this person.'" He specified that the feature would only be available while wearing the glasses and would surface the name of a person the user had previously met.

Bosworth later said of NameTag, "I think would be a great feature." Meta spokesperson Ryan Daniels emphasized to WIRED that Bosworth used the conditional "would," bolding and underlining the word in an email exchange. Daniels stated there was "no contradiction" and that the company was "exploring" the feature, which is "not available to consumers today." He also stressed that NameTag remains distinct from connecting glasses to a central database of people, which Meta says it is not building.

How the System Works—and the Legal Stakes

WIRED's analysis of the Meta AI app found that NameTag converts faces captured by Meta glasses into unique numerical signatures known as "faceprints." These faceprints could be compared against a face database stored on users' own devices, which was populated by Meta's servers. Bosworth asserted on the podcast that NameTag will not rely on a central database—a point neither Thompson nor WIRED had claimed or suggested it would.

The distinction between a central database and millions of local databases connected to Meta's servers carries significant legal weight. State laws such as Illinois' Biometric Information Privacy Act (BIPA) and Texas' Capture or Use of Biometric Identifier Act (CUBI) restrict companies from publicly deploying face-recognition technology without explicit user consent and other safeguards.

Meta has faced these issues before. In 2019, the company abandoned its "Tag Suggestions" automatic face-recognition feature on Facebook following a $5 billion settlement with the Federal Trade Commission over privacy issues including face recognition, and before a $650 million settlement with the state of Illinois specifically over Tag Suggestions.

By keeping face-matching on the user's device rather than querying a central server, Meta may be positioning itself to argue compliance with laws like BIPA and CUBI. Whether that approach is legally defensible remains contested. In 2021, a federal judge allowed a BIPA class action lawsuit against Apple's Photos app to proceed, ruling that a company could plausibly "possess" faceprints stored on users' own devices. That case was certified as a class action in June 2026 and awaits rulings on several motions. However, an Illinois appellate panel ruled in 2022 that Apple did not "possess" Face ID data stored solely on devices, and a federal judge dismissed a similar suit against Samsung's photo app in 2024 because Samsung never received or accessed the face data its software generated.

The dividing line in these cases has hinged less on where data resides and more on who controls it—whether the feature is optional, whether users can disable it, and whether the company can ever access the data. When WIRED asked Meta in June whether NameTag would be opt-in and how the system retained faceprints and cropped images, the company declined to answer. It also would not explain why it licensed a third party's face-recognition software, when that arrangement began, or whether it continues.

Meta has designed and built a face-recognition system, placed it on millions of phones, and had a prominent executive praise it on a public platform. Whether that constitutes a feature that "exists" may depend on who is asking—and who is answering. If this debate over the semantics of existence versus deployment caught your attention, share this article with your network and join the conversation about the future of face recognition in everyday wearables.

Source: Wired