June 4, 2026
Glasses, but make it creepy
Meta's ships facial recognition on smart glasses
People saw a hidden face-ID system and instantly screamed 'surveillance glasses!'
TLDR: A researcher found Meta’s smart-glasses app includes a hidden, working face-recognition setup, though there’s no proof it’s active for normal users. Commenters immediately turned this into a privacy firestorm, calling the glasses creepy surveillance gear and demanding regulators step in.
The big reveal here isn’t that Meta’s smart glasses are secretly naming strangers today — it’s that a researcher says the app already contains the parts needed to do it, sitting there quietly behind the scenes. In plain English: the glasses app appears to include the tools to spot a face, turn it into a unique digital signature, compare it with a stored list, and even pop up a “Person Recognized” alert. The researcher says they could trigger it on a test image, but also stresses there’s no proof this is turned on for ordinary users right now. That nuance, however, did not stop the comment section from going full panic mode.
Readers came in swinging with words like “creepy,” “disgusting,” and “portable surveillance cameras on their face.” One commenter basically summed up the mood as: even if privacy-conscious people opt out, it won’t matter if everyone else starts wearing always-on camera glasses in public. Another went straight for the legal drama, hoping Illinois — famous for strict biometric privacy law — would sue Meta into oblivion. But the spiciest twist came from someone claiming to be a former Facebook researcher, who said legal teams and even US regulators had historically pushed back hard on face recognition. That turned the thread into a mini courtroom: is this a sinister plan waiting for launch, or just dormant code from an internal experiment? Either way, the internet has already delivered its verdict: this is dystopian, and deeply uncool.
Key Points
- •The article says Meta’s `com.facebook.stella` Android app version `273.0.0.21` contains a complete on-device facial-recognition stack, including models, storage, indexing, and notification logic.
- •The author reports successfully running the pipeline on a test image by invoking an existing handler, producing a 2048-dimensional face embedding and triggering a "Person Recognized" notification.
- •The article says three ExecuTorch models are delivered through Meta’s NMLML system: SCRFD for face detection, KPSAligner for face alignment, and SFace for biometric embedding.
- •The author states there is no evidence from the test account that the feature is enabled for ordinary users, noting the UI was absent and no identity data was observed being pushed from Meta servers.
- •The article identifies a local database path used by the recognition pipeline and concludes that the capability appears assembled on-device but gated by Meta.