What Was Discovered Inside Meta’s Smart Glasses App?
Meta smart glasses privacy concerns center on a hidden face recognition system embedded in the companion Meta AI app, where undisclosed code can detect faces, create biometric identifiers, and store these data points without the clear, informed consent users would reasonably expect from consumer hardware. Reverse engineering of the app, internally called Stella, shows three machine learning models working together to detect faces, align them, and generate 2048‑dimensional biometric fingerprints. According to Gadget Review, the app even includes an SQLite vector database for similarity matching, persistent storage for unknown faces, and a notification pipeline that can announce “Person recognized” when it finds a match. These capabilities sit behind a dormant feature called NameTag, later rebranded as “Connections,” which promises to “remember the people you met” but is not visible to normal users today.
How the Hidden Face Recognition Technology Works
The face recognition technology goes far beyond basic camera autofocus. Meta ships three ExecuTorch models in the Meta AI glasses app: SCRFD for face detection, KPSAligner for positioning, and a scaled-up SFace variant to turn each face into a numerical embedding, or faceprint. These faceprints are stored in a local vector database that supports fast similarity search. If activated, the system would compare each captured face against entries stored on the user’s phone and trigger an on-device “Person recognized” notification when it finds a match. Unrecognized faces are not discarded; they are saved as cropped images plus binary embedding files in a private directory that survives device reboots, effectively building a biometric data collection pipeline of “faces pending identification.” During tests by Buchodi researchers, preloading a portrait of Michel Foucault allowed the system to identify him end-to-end, proving the chain works in practice.
Consent, Transparency, and Meta’s History With Faceprints
The discovery lands against a sensitive backdrop for Meta smart glasses privacy. Back in 2010, Facebook deployed a large-scale face recognition system that scanned billions of photos and later cost the company over $2 billion in settlements before it was shut down in 2021, along with more than a billion stored faceprints. Now, internal documents reported by WIRED describe plans to roll out NameTag during a “dynamic political environment,” while Meta publicly claims that “nothing has shipped to consumers, and no final decision has been made.” Privacy advocates, including more than 70 organizations such as the ACLU and Fight for the Future, have already urged Meta to abandon face recognition technology in its smart glasses. Their concern is that quietly shipping hidden surveillance code in mass-market wearables revives unresolved questions about consent, retention, and how biometric data might be handled over time.
Why This Matters for Everyday Users
For everyday users, the troubling part is not only the sophistication of the face recognition technology but its invisibility. The app has no active “Connections” interface, no clear controls, and no up-front explanation that biometric data collection might occur through smart glasses they wear throughout the day. Meta’s own transparency materials indicate AI glasses data can be stored locally and on remote servers, which raises questions about where faceprints might travel if the feature is enabled later. Unlike taking a deliberate smartphone photo, smart glasses can watch continuously, turning routine interactions at work, in shops, or on the street into a stream of potential biometric identifiers. Cooper Quintin of the Electronic Frontier Foundation warns that embedding hidden surveillance code into popular devices risks turning customers “into a distributed surveillance machine” while leaving bystanders unaware they are being scanned.
Practical Steps Users Can Take Now
While Meta says no face recognition feature has launched, users concerned about Meta smart glasses privacy can still reduce risk. First, review app permissions for the Meta AI companion app and limit camera and background access where possible, especially when you are not actively using the glasses. Second, turn off continuous capture modes on the glasses and avoid wearing them in private or sensitive settings where bystanders cannot meaningfully consent. Third, follow Meta’s official policy updates and changelogs closely; the appearance of a “Connections” or similar feature should prompt a fresh review of settings and data export tools. Finally, consider whether you are comfortable with any level of biometric data collection in consumer wearables. The hidden surveillance code in this app shows that AI capabilities can arrive long before clear, transparent choices are offered to the people affected.
