On June 4, WIRED published a code audit report, revealing that Meta has built a facial recognition system codenamed “NameTag” for its AI glasses. The system uses the cameras on the smart glasses to capture facial images and match them with biometric data stored locally on users’ phones to complete identification. Just one day after the report went live, Meta rushed to roll out an app update and fully removed all relevant code. This dramatic sequence of events has once again landed Meta in the eye of a privacy controversy.
Code Already Integrated, Yet Meta Claims the Feature Is Still Under Consideration
WIRED’s investigation found that Meta integrated the core components of NameTag into the Meta AI app as early as January this year. This app, boasting over 50 million downloads, serves as the essential control software for Ray-Ban Meta, Oakley Meta and other Meta-powered AI glasses. Three AI models supporting the feature have been delivered from Meta’s servers to users’ mobile devices: one for face detection, one for image cropping, and the third for processing biometric data.
Nevertheless, Meta repeatedly stated publicly that the technology was still under review. A Meta spokesperson remarked back in April that the company would conduct thorough assessments even if it planned to launch the feature in the future. The stark discrepancy between the embedded code and public statements has put Meta in an awkward contradictory position.
Recognizing Known Faces and Storing Unidentified Ones
Code analysis shows that the operational workflow of NameTag is fully developed. Faces captured by the AI glasses are converted into biometric feature codes and cross-referenced against the local database on the user’s phone. The system sends out a notification once a match is found. For unrecognized faces, the images will be cropped, indexed and saved to a dedicated pending folder.
In app versions released in May, the feature was rebranded as “Connections”, framed as a tool to help users keep track of people they meet. Independent cybersecurity researchers have replicated and verified these findings.
Cooper Quintin, a researcher at the Electronic Frontier Foundation’s Threat Lab, noted that while the feature is not yet available to general users, it is nearly complete. Independent researcher Buchodi ran a test by importing the facial feature data of the late French philosopher Michel Foucault, and the system successfully popped up an identification alert. “All core modules for facial recognition have been installed in Meta’s companion app; the feature is just a few steps away from official launch,” Buchodi commented.
A Troubled Past: Billions of Facial Templates and Over $2 Billion in Settlements
This is far from Meta’s first run-in with scandals over facial recognition technology. Back in 2010, Facebook launched its automatic photo tagging feature, which soon amassed data from over a billion users and became one of the world’s largest civilian facial recognition systems.
In 2019, Meta paid a $5 billion settlement to the US Federal Trade Commission over privacy violations including its facial recognition practices. In November 2021, Meta announced it would shut down its facial recognition service and delete more than one billion facial templates. Lawsuits against the company, however, did not end.
Meta reached a $650 million settlement for a class-action lawsuit in Illinois, followed by a $1.4 billion settlement over biometric data disputes in Texas in 2024. Combined, Meta has paid more than 2 billion US dollars in penalties for its past use of facial recognition.
Joseph Jerome, who joined Meta’s Reality Labs in 2021 to oversee AR and VR privacy reviews, revealed that the company never regarded the 2021 shutdown as a permanent move. “Teams across the company have constantly discussed when to bring facial recognition back.”
70+ Advocacy Groups Speak Out, Meta Rushes to Delete Code
In April, more than 70 advocacy groups including the American Civil Liberties Union (ACLU) and the Electronic Privacy Information Center jointly urged Meta to scrap the NameTag project, warning that the technology could enable stalking, harassment and other abusive behaviors.
Following WIRED’s exposé, Meta acted with unusual haste. Within 24 hours, it pushed an update to the Meta AI app, removing core facial recognition modules, the entire NameTag identification process, “Person recognized” notification prompts, and local cache directories used to store snapshots of unrecognized faces.
Meta spokesperson Andy Stone insisted the feature was merely in the exploration phase and no final launch decision had been made. Kade Crockford, director of the Technology for Liberty Program at the Massachusetts ACLU, pointed out that deleting code after the fact does not erase the reality that Meta attempted to roll out the feature covertly.
From Aids for the Visually Impaired to Normalized Surveillance
Facial recognition does have legitimate use cases. A 2018 joint study by researchers from Cornell Tech and Facebook found that visually impaired participants considered facial identification an essential function for daily mobility, and specialized assistive devices already offer such features with pre-registered contacts.
Privacy advocates, however, fear the technology will evolve in problematic directions. Woodrow Hartzog, a professor of privacy law at Boston University, argued that user consent mechanisms are fundamentally insufficient. In many cases, permission requests are tied to employment benefits or access to essential services. Companies frame privacy trade-offs as personal choices, allowing unrestricted data collection while claiming users retain full control.
“The more widely this technology is deployed, the more people will grow to see it as ordinary,” Hartzog said. “Once facial scanning becomes commonplace, public moral judgments about the technology will shift.”
Jerome, the former Meta employee, put it more bluntly: embedding such technology into Meta’s product ecosystem is essentially setting industry standards. He expressed doubts over whether Meta could ever implement the technology in a fully compliant and responsible manner.
Code Is Gone, But Core Problems Remain
Meta’s last-minute code removal is widely seen as a crisis control tactic rather than a genuine reflection on its practices. The core issue surrounding NameTag lies not in the existence of code, but in Meta’s attitude toward facial recognition. After publicly deleting over a billion pieces of facial data in 2021, the company quietly embedded a new facial recognition tool into an app used by tens of millions just four years later.
Cameras on consumer-grade AI glasses stay active most of the time, capturing faces passively and unobtrusively. Unlike smartphone photography, wearers’ glasses can record bystanders without their knowledge or consent.
Meta claims it will not build a centralized facial database. According to the code, though, the system pulls facial feature codes from Meta’s servers and stores them locally on individual devices. This distributed storage model does not eliminate inherent risks. The technological capability has already been deployed across tens of millions of devices, and the feature could be reactivated with just a few lines of code.
About the Author
