Google’s Android XR Glasses Blur Faces Before Gemini Sees Them. Here’s the Catch
Android XR glasses blur faces before Gemini sees them and ban face-ID apps. Here's what Google's privacy rules cover, and what they miss.
In this article
Smart glasses have a people problem, and it isn’t the people wearing them. It’s everyone else in the room. Google clearly knows that. With its first Android XR glasses now up for pre-order at $395 and shipping December 10 in the U.S. and South Korea, the company has laid out a set of camera rules that go further than anything Meta has shipped: faces get blurred before Gemini ever sees an image, and third-party apps on Google Play are barred from running real-time facial recognition.
That’s a good start. It is not the same as a guarantee, and the gap between the two is worth understanding before you put a camera on your face, or sit across from someone who has.
How the face blurring works
When you ask Gemini about something in front of you, the glasses’ 12MP ultrawide camera grabs a frame and sends it to your paired phone. Before that image goes anywhere near Google’s servers, the phone blurs every face in it. That includes bystanders, and it includes you, if you happen to be looking in a mirror.
Google’s pitch, from Google’s Shahram Izadi as quoted by Android Police, is simple:
“Personal help doesn’t come at the expense of anyone’s privacy.”
A few other promises ride along with the blur. Images captured for Gemini requests are deleted after use, don’t show up in your Gemini chat history, and aren’t used for AI training or ad targeting, according to 9to5Google. Voice data sent to Gemini is filtered to keep only the wearer’s voice, and live translation listens within a 60-degree cone in front of you rather than to the whole room. Gemini is also supposed to use the camera only when a request clearly calls for it; vague prompts get a follow-up question first.
The rules Google is putting on apps and wearers
The more interesting part may be the policy layer, because that’s what applies to everyone who isn’t Google.
- No face ID apps: Google Play apps can’t use the glasses’ camera for real-time facial identification or social live-streaming.
- Hands off the light: Developers can’t change how the outward-facing recording light behaves. If the light is covered or fails its regular self-check, the camera won’t record.
- Short clips: Video recordings top out at 90 seconds. 9to5Google notes Meta’s glasses allow up to five minutes.
- No light-blocking accessories: Google’s policies ban selling or advertising stickers and add-ons that hide the indicator.
- User agreement: Wearers must agree not to record people without permission, not to record in sensitive places and not to cover the light. Google says violations can cost you the ability to use the device.
- Adults only: Activating the glasses requires a Google account for someone 18 or older.
There’s also a physical switch on every pair to turn the glasses off entirely, which is the kind of low-tech control that privacy advocates have asked for since Google Glass.
What the rules don’t cover
Here’s the catch, and it’s a big one. The recording light only comes on when the glasses capture photos or video you can view later. When Gemini uses the camera for context, the light stays off. Google’s logic is that those frames are blurred and thrown away. From the point of view of the person across the table, though, a camera is looking at them and nothing says so.
Biometric Update raised a second problem: it’s not clear whether the facial recognition ban is enforced by Android XR itself or just by Play Store policy. Third-party apps reach the camera through standard Android interfaces, and Google’s developer documentation doesn’t describe an operating-system control that would stop an approved app from running face identification anyway. Apps installed from outside Google Play are an open question too. None of that proves anyone can actually build a face-scanning app for these glasses. It does mean the protection, as documented so far, rests on Google catching bad actors rather than making the feature impossible.
Google also hasn’t claimed the on-phone blurring is perfect, so it’s possible, at least in theory, for an unblurred face to slip through to its servers.
Why Google is going first on this
Timing matters. Meta, the market leader, is under fresh scrutiny: on October 29, Senators Edward Markey and Ron Wyden sent the company questions about AI training, indicator lights and camera-based AI features, with a response due November 25. Meta has explored facial recognition for its glasses but says no such feature has shipped. Across the Atlantic, an Irish parliamentary committee recommended on November 10 that glasses capable of covert recording be banned, and European regulators have urged makers to blur bystanders and make AI camera use visible.
Against that backdrop, Google is trying to make privacy a selling point rather than a liability. For a product that also has to win over people who will never buy it, that’s smart positioning.
What this means if you’re buying a pair
If you pre-ordered, or you’re tempted at $395, the practical takeaways are short. You’ll need a phone to pair with: Pixel or Galaxy devices on Android 17 or One UI 9, or an iPhone. Battery life is rated at up to nine hours. Expect the 90-second video cap to feel limiting if you planned to film whole concerts. And expect to be asked about the camera, a lot.
The fair verdict: Google’s rules are the strongest any mainstream smart glasses have launched with, and the face blurring is a real technical commitment, not just a policy page. But the Gemini-without-a-light gap is the one I’d want closed before these become as common as earbuds. A camera that’s quietly looking should still say so.
Sources: 9to5Google, Android Police, Biometric Update. Photo by Zarak Khan on Unsplash.


