Meta Embeds Unreleased Face-Recognition Code in Smart Glasses App
WIRED analysis finds Meta has integrated biometric identification software into its companion app, contradicting statements that the technology was still under review.
An analysis by WIRED has revealed that Meta has quietly embedded unreleased face-recognition code into its Meta AI companion app, a platform downloaded by more than 50 million users. The system, internally designated as “NameTag,” is designed for Meta’s Ray-Ban and Oakley smart glasses and converts facial images captured by the device’s camera into biometric signatures. These signatures are then checked against data stored on the user’s phone, a process that has been integrated into software distributed since January.
The discovery contradicts public statements made by the company in April, when Meta indicated it was still evaluating the technology and had not made a final decision on its deployment. While the feature is not currently enabled for consumers, core components have been present in the live app for months. A version of the app released in May rebranded the function for users as “Connections,” encouraging them to “remember the people you met,” suggesting the interface was prepared for activation.
Three artificial intelligence models powering the NameTag system have been deployed from Meta’s servers to customer devices. According to the code review, one model detects faces, a second crops them, and a third encodes them into biometric data. Only traces of the user interface remain, but the underlying infrastructure is in place. If activated, the system would trigger notifications when a face is recognised, while unrecognised images are cropped, indexed, and saved to a folder marked as pending.
Privacy advocates and security researchers have strongly criticised the move, warning that it could normalise surveillance and facilitate stalking. The Electronic Frontier Foundation and the American Civil Liberties Union were among more than 70 groups that previously demanded Meta scrap the feature. Independent researchers, including Cooper Quintin of the Electronic Frontier Foundation and an independent expert known as Buchodi, reproduced key aspects of the analysis. Buchodi tested the pipeline by adding a faceprint of the late French philosopher Michel Foucault, which successfully triggered a “Person recognised” notification.
Meta maintains that it is not building a central face database and that the system is designed to pull faceprints from its servers and store them on user devices. The company stated that nothing has shipped to consumers and no final decision has been made regarding a public rollout. However, the integration of this technology revives a capability Meta said it had sunsetted in 2021, following significant legal settlements over biometric data collection. The move raises questions about consumer consent and the potential for widespread biometric identification in public spaces.
