Google Puts Privacy Switches on Its Smart Glasses, Dares Rivals to Match

Google will start selling its first Android XR Intelligent Eyewear smart glasses on December 10, and it is shipping them with an unusual pitch: a stack of privacy controls designed to persuade the public that camera-equipped eyewear is not inherently creepy.

The glasses, built in collaboration with Samsung, Warby Parker, and Gentle Monster, are available for preorder now. They enter a market where Meta's Ray-Ban glasses have earned wearers the label "Meta Creeps" — an echo of "Glasshole," the term coined when Google Glass launched in 2013. The perception problem is concrete: the cameras are hard to see, the recording indicator light is less conspicuous than a raised smartphone, and people have tampered with the LEDs, blocked them, and recorded women without consent.

Google says it has decided to treat privacy as a design constraint rather than an afterthought. "There's this whole feature set of AI privacy that we think is super important," Juston Payne, Google's senior director of product management for XR, told WIRED. "Ultimately, what we're trying to do here is say: Look, these are choices. These are, honestly, not hard technical problems. They're decisions. And we would like other companies to make different decisions, sort of like a bar-raising thing."

The stakes extend beyond optics. Smart glasses are data-collection devices worn on the face, and privacy researchers warn they are a precursor to mass surveillance with the capacity to identify people in public. This summer, Meta silently added a face-recognition code to its smart glasses and deleted it only after a WIRED report; in 2024, two Harvard students built a face-recognition tool for Meta's glasses.

What hardware controls do the glasses ship with?

Every pair in the Intelligent Eyewear line includes a power switch that turns the glasses off completely, so users can wear them as ordinary glasses. A recording light illuminates whenever the glasses snap a photo or record a video, and recording is intentionally limited to 90 seconds for privacy purposes.

The glasses also carry tamper protection. The camera detects in real time whether a user is covering the indicator light while taking a photo or recording, or whether the LED has been tampered with. In either case, the glasses refuse to capture the image or video. Payne says Google's algorithms always check whether the LED is functioning properly, and while "it's not perfect," the company is satisfied with its performance so far.

How does Google handle the camera data itself?

When a user asks Gemini to look at something — translate a menu, or add concert details from a poster to a calendar — Google does not save that visual data in Gemini History. The company promises not to use it for AI training or ad targeting. The data is encrypted on the devices, and Google says it is encrypted in transit whenever it travels to the cloud or to a phone for processing. Google also plans to bring its Private AI Compute technology, comparable to Apple's Private Cloud Compute, to process this data in a secure environment.

Third-party apps can access the camera, but they must request permission first, and users can revoke access at any time. The Google Play Store will prohibit apps from using the camera for real-time face recognition, social livestreaming, or altering the recording light's behavior.

What protects the bystanders?

Two pipeline features target people near the wearer rather than the wearer. When Gemini needs the camera to answer a question about the user's surroundings, the image-processing pipeline redacts faces in the background before Gemini processes the visuals. "That is designed to get around any kind of face recognition," Payne says.

Audio gets similar treatment. The glasses use directional microphones that, by default, pick up only the wearer's voice, and a "denoiser step" in the pipeline strips out nearby conversations before audio reaches Gemini.

Google also binds the hardware to contractual rules. During setup, users must agree not to record others without permission, not to record in sensitive places, and not to cover the recording indicator light. Violations breach Google's terms of service, and "a user's ability to use the glasses may be disabled." Setup requires an 18+ Google account.

The policies extend across Google's commercial properties. Google Shopping and Google Ads will ban the sale and advertising of stickers or accessories designed to hide recording lights. On YouTube, content restrictions already prohibit videos of harassment and unwanted sexualization, regardless of whether they were recorded on glasses, cameras, or smartphones — but Payne says Google is making sure the YouTube team is aware and ready to enforce the policy as the smart glasses platform launches.

Why is Google publishing these policies now?

Payne frames the measures as an industry challenge, not just a product feature. "We hope that never again do glasses launch from some other company that doesn't have good algorithms built in," he says. "That should not have happened; hopefully it doesn't happen again."

He argues the industry owes the public directness after years of eroded trust. "Ambiguity is not your friend here," Payne says. "I think that a lot of that concern came from some trust being broken. So there is a rebuilding thing happening right now … You can hopefully have your cake and eat it too, which is cool glasses that everybody's excited about—that you're comfortable with."

What do privacy experts say?

The experts WIRED spoke with credited Google with avoiding Meta's mistakes but stopped short of endorsing camera glasses as normal streetwear. Even with an indicator light, they noted, it remains harder to tell when someone is recording you in public than when someone holds up a smartphone.

Reem Suleiman, senior campaign director at the digital rights group Fight for the Future, argues camera-equipped glasses should not be normalized because they are a precursor to mass surveillance and can identify people. "People find ways around it," Suleiman says. "There's no actual aspect of consent."

Janusz Swierczynski, who researches human interactions with emerging technology at the University of Oxford's Saïd Business School, calls the privacy measures meaningful but says they address privacy at the individual level, largely ignoring the bystanders affected by the devices. He attributes that gap to technology advancing faster than policy and research.

"A lot of privacy is managed by social norms, by what we think is acceptable to record, and that prevents people from recording anything and everything," Swierczynski says. "Normalizing recording does have unintended consequences, which is why it's important to focus on bystander privacy."

The December 10 launch will test whether engineering choices — a 90-second cap, tamper detection, face redaction, denoised audio — can outweigh a decade of accumulated suspicion. Payne is betting that rivals who skip those safeguards will now face an explicit comparison.