Does Meta Blur Faces in Smart Glasses Videos? [2026]

Meta has told users, regulators, and reporters that it blurs faces in the footage its Ray-Ban and Oakley smart glasses send off for human review. That claim is now the center of a class action lawsuit, and independent reporting says the blur Meta describes doesn't work the way the company implies. If you wear the glasses, appear in someone else's footage, or just want to know whether a device's built-in privacy promise is something you can rely on, the honest answer is: not yet, and maybe not ever without a step you control yourself.

This guide walks through what's actually confirmed about Meta's face-blurring claims, the lawsuit and reporting behind it, the facial recognition feature Meta quietly built and then pulled after being caught, and what genuinely protects the people who end up in someone's point-of-view footage — including yours.

TL;DR: What's Confirmed So Far

Does Meta Actually Blur Faces in Smart Glasses Footage?

Meta has said, in various forms, that faces are blurred before footage reaches the humans who review it for AI training. The claim doesn't hold up consistently under scrutiny. The Electronic Frontier Foundation examined Meta's practices and concluded plainly that "it seems like the company tries to blur faces automatically in footage it captures for annotation, though it's not always reliable" — see EFF's full analysis of Meta's Ray-Bans.

The March 2026 class action against Meta makes the same point with more legal weight. According to the Euronews report on the lawsuit, Meta claims faces are "usually blurred in images," but sources indicate that protection doesn't work consistently. "Usually" is doing a lot of work in that sentence — it's an admission, not a guarantee, and the gap between the two is exactly where the lawsuit lives.

It's worth being precise about what this blur even applies to. It's not a feature you turn on for your own saved clips. It's an internal step Meta says it takes before footage goes to contractors for annotation — a much narrower promise than "your footage is anonymized," and one the company itself frames with hedging language.

What Did the Meta Smart Glasses Lawsuit Actually Allege?

Two plaintiffs, Mateo Canu of California and Gina Bartone of New Jersey, filed suit against Meta and Luxottica of America — the glasses' manufacturing partner — represented by the Clarkson Law Firm. The core allegation is false advertising: Meta marketed the glasses as "designed for privacy, controlled by you" while, according to the complaint, routing user footage to overseas contractors for AI model training without adequately disclosing that fact.

The lawsuit names three specific harms. First, footage containing nudity, intimate moments, and even financial documents was reviewed by people the wearer never consented to share it with. Second, the automatic blur Meta pointed to as a safeguard reportedly fails inconsistently. Third, the marketing language created an expectation of privacy the product didn't deliver.

This isn't the first time Meta has paid for facial data practices — the company settled Illinois Biometric Information Privacy Act claims for $650 million and a similar Texas case for $1.4 billion in 2024, over its earlier Tag Suggestions feature. A pattern of large biometric settlements is relevant context for how seriously to weigh a new claim that a privacy safeguard doesn't work as described.

Who Were the Human Reviewers, and What Did They See?

The footage in question didn't stay inside Meta. It went to Sama, a Nairobi, Kenya-based data annotation company that Meta contracted to help label footage for AI training. Workers there reviewed clips captured through the glasses' first-person camera, and the material wasn't limited to street scenes or public spaces.

One Sama worker's description, cited in reporting on the case, was blunt: "We see everything — from living rooms to naked bodies." That's the practical consequence of a wearable camera that captures whatever the wearer happens to be looking at, uploaded to the cloud by default, and then handed to contractors on another continent for review — with a face-blur step in between that the company itself only claims works "usually."

For a deeper look at how this fits into the broader pattern of smart glasses being used to record people without consent — including the BBC investigation into covert filming that first surfaced the Sama contract — see our guide on smart glasses covert filming and how automated redaction protects bystanders.

Did Meta Secretly Build Facial Recognition Into the Glasses?

This is the part that turns "unreliable blur" from an embarrassing bug into a pattern. In June 2026, WIRED reported finding a dormant facial recognition system built into Meta's AI companion app — internally called NameTag — that had already been distributed to roughly 50 million app users before anyone outside the company noticed. According to BiometricUpdate's coverage of the WIRED report and the patent that later reignited the debate, the system was designed to convert faces captured by the glasses into biometric signatures and compare them against a database stored on the user's own device.

Meta removed most of the NameTag components shortly after WIRED's story ran. The company hadn't announced the feature, hadn't sought consent for it, and only pulled it back after external reporting forced the issue — the same pattern as the annotation-blur claim: a privacy safeguard exists on paper, but only external scrutiny confirms whether it actually behaves the way the company says.

Three US senators — Ed Markey, Ron Wyden, and Jeff Merkley — sent Meta a letter demanding transparency on the facial recognition plans, asking how the company would obtain consent, handle biometric data, test for bias, and prevent misuse, with an April 6, 2026 deadline for a response. We cover that letter and the broader ACLU-led coalition response in more detail in our guide to the Meta smart glasses facial recognition backlash.

Northeastern University researchers put the stakes plainly: facial recognition on a stranger could hand a wearer "a full dossier" on someone they've never met, built from a face alone. EFF senior staff attorney Adam Schwartz added that "there are not enough limitations on what the companies are doing with that data" — see Northeastern Global News' full report.

Is It Safe to Use — or Be Filmed by — Meta Smart Glasses?

The honest answer splits into two different questions, and conflating them is where a lot of coverage goes wrong.

If you're the wearer: the hardware works as advertised, but the privacy layer around your own footage has real, documented gaps — cloud upload on by default, AI conversation audio retained by default, and a recording light that's small enough to miss and can reportedly be disabled outright in some modes. None of that is illegal, but it's not what "designed for privacy, controlled by you" implies either.

If you're a bystander: you have almost no visibility into whether you're being recorded, and once you are, you have no control over where that footage goes. The recording indicator is the only signal you get, and multiple venues — including Soho House, Wetherspoon pubs, and various UK theatres — have concluded it isn't reliable enough to trust, banning the glasses outright rather than relying on it.

Neither of these problems is solved by Meta's internal annotation blur, because that blur only applies to a narrow slice of footage Meta routes for its own training purposes — not to what gets saved, exported, or shared by the wearer.

How to Blur Faces in Your Own Footage — The Step That's Actually in Your Control

Whether you're a smart glasses wearer trying to publish footage responsibly, or you're editing bodycam, dashcam, or vlogging clips that happened to catch bystanders, the reliable version of "blur their face" is the one you run yourself before the clip goes anywhere.

Step 1: Upload Your Clip

Go to bgblur.com and drag in your video — MP4, MOV, or M4V, up to 4K, including the format smart glasses export by default.

Step 2: Let AI Detect Every Face

BGBlur's motion-tracked detection finds every face in the frame and follows it as people move through the shot, so a bystander walking across the scene stays blurred for the full clip rather than just the frame where they were first spotted.

Step 3: Review and Export

Preview the result, adjust if needed, and export as MP4, MOV, or WebM. Your upload is processed and deleted within 24 hours — no human reviewer, no overseas contractor, no separate database of your footage sitting on someone else's servers.

This same workflow works for blurring faces in photos and video in about three seconds, for license plates and other identifying details, and it's the same standard we walk through for GDPR-compliant video publishing if you're handling footage in a professional or regulated context.

Meta's Blur vs. Blurring It Yourself

Honest Limitations

Face blur — Meta's or anyone else's — cannot stop live facial recognition running on someone else's device. If a wearer's glasses run identification in real time, that happens before a file even exists, and no downstream editing tool changes that. Blurring is the right tool for controlling what happens to footage after it's captured and before it's shared — not for preventing on-device identification you have no access to. For very large archives, like a public agency processing years of bodycam footage, a dedicated enterprise redaction pipeline may be a better fit than a per-clip web tool; BGBlur is built for the much more common case of an individual clip or batch you want blurred before you post it.

The Bottom Line

Meta's own account of its face-blurring practices doesn't survive contact with a class action lawsuit or independent reporting. The company says faces are "usually blurred," a phrase that concedes the failure mode it's now being sued over, and the same reporting cycle that surfaced that inconsistency also caught Meta quietly building — and then pulling back — a facial recognition feature nobody had agreed to. The pattern across both stories is the same: a privacy claim that only holds up as long as no one checks.

The fix isn't waiting for Meta to get its internal blur right. It's not routing your footage through a system where the safeguard is optional and the review is outsourced. If you're publishing video that includes other people — whether it came from smart glasses, a bodycam, or your phone — blur the faces yourself with BGBlur before it ever leaves your hands.