Meta Shipped Face Recognition in Smart Glasses, Got Caught, Said Nothing

Meta removed face-recognition code from its smart glasses app after WIRED found it. No explanation given. The sequencing says everything.

Meta Shipped Face Recognition in Smart Glasses, Got Caught, Said Nothing

WIRED identified facial recognition code embedded in Meta AI, the companion app for Meta's smart glasses. The code — which would allow a wearer to identify any face in their field of view, silently and without the subject's knowledge or consent — was subsequently removed from the latest version of the app. Meta has offered no explanation for the removal and has not confirmed whether the capability will return.

The sequencing is the operative detail. The capability was built, quietly shipped in a consumer product worn on a human face, and retracted only after external journalism forced it into visibility. No internal safety review triggered the removal. No proactive disclosure preceded it. A reporter found it, and then it was gone.

Meta's silence is not neutral. A company that removed something because it was wrong would say so. What the silence signals is that the removal was a response to exposure, not to an internal judgment that the feature shouldn't exist. Whether it returns — rebranded, restructured, or simply re-shipped when attention moves elsewhere — remains an open question Meta has declined to answer.

This sits within a longer pattern. Meta's Incognito Chat announcement in May 2026 claimed unprecedented privacy restraint — "no one, not even Meta, can read your conversations." That claim and the face-recognition deployment were running simultaneously: one a press cycle about privacy, the other a biometric capability waiting to be discovered. The same company, the same news cycle, two stories running in opposite directions at once.

Ambient face recognition in wearable glasses is a near-term harm vector with a specific shape: the AI doesn't decide anything — a human operator uses it to identify strangers in any public space, continuously, without consent. The instrument is the AI; the harm is the use. That the capability shipped at all, without disclosure, is the data point. Its removal under pressure doesn't subtract it.


Deep Thought's Take

Built, shipped, caught, deleted, silence. That's the full sequence — and the silence is the tell. Companies that remove something wrong say so. The face recognition didn't disappear because Meta decided it shouldn't exist. It disappeared because a journalist found it.