A privacy fight that played out in roughly 24 hours says a lot about where always-on wearables are headed. On June 4, 2026, WIRED reported that Meta’s AI companion app — the one that pairs with its Ray-Ban smart glasses — contained a hidden, dormant facial-recognition feature internally called “NameTag.” By June 5, Meta had stripped the code out. The speed of the deletion is the tell.

What was actually in the app

According to the reporting (EFF, The Next Web, Gizmodo), researchers found more than a stray reference. NameTag reportedly included face-detection models, biometric matching tools, local databases, and alerting — the components of a working system to identify a person seen through the glasses in real time. Crucially, this code shipped inside an app that has been downloaded onto more than 50 million devices.

Meta’s defense, delivered loudly, is that the feature was never enabled and was “exploratory.” A company spokesman accused WIRED of burying that detail. Both things can be true at once: the feature was off, and it was built, packaged, and distributed at scale.

Why “dormant” is not “harmless”

The instinct to wave this away — “it wasn’t even turned on” — misses how software risk works. A capability that is fully built and shipped is one configuration flag away from being live. The hard engineering — the models, the matching pipeline, the local store — is the part that takes months. Toggling it on is the easy part.

So the meaningful facts are: the system existed, it was complete enough to run, and it was inside 50 million installs. That it was disabled is a policy choice, and policy choices are reversible without warning. The 24-hour delete after public exposure underlines the point — this was governed by reputational pressure, not by a technical impossibility.

The real problem: bystanders can’t consent

Set aside the corporate back-and-forth and the core issue is structural. Face-ID built into camera glasses breaks the one privacy principle that matters most in public space: consent. The wearer might agree to terms; the stranger on the sidewalk who gets silently identified, named, and looked up never did — and usually has no idea it happened. Privacy advocates have made this argument for years, and it is exactly why a latent NameTag alarmed people more than a clearly-labeled feature would: covert identification removes the bystander’s ability to even object.

This is not hypothetical. In 2024, students demonstrated that off-the-shelf Meta glasses plus public face-search tools could de-anonymize strangers in real time. NameTag would have folded that capability into the official app.

Bottom line

Meta deleted the code, and the EFF and others fairly called it a win — public scrutiny worked. But the durable lesson is not “Meta backed down.” It is that the surveillance capability is being quietly built into mainstream consumer hardware, switched off by policy rather than absent by design. For anyone thinking about privacy in public, the question is no longer whether the glasses can identify you — increasingly they can — but who gets to flip the switch, and whether you will ever be told when they do.

Covering privacy and surveillance tech — corrections or tips welcome via @mrtdnet on Telegram.