Sunday, November 15, 2026 Independent reporting on artificial intelligence Newsletter
AI News Beats
Policy & Society

Google spells out camera privacy rules for Gemini eyewear

Google says the glasses will signal photo and video capture and that camera visuals used by Gemini will not be saved in Gemini History or used to train AI.

A pair of eyeglasses rests beside a notebook and laptop
Photo: 2H Media / Unsplash

Google described privacy protections for eyewear that uses its Gemini assistant on November 12, setting out how it plans to handle camera data and notify people when a wearer records. The design puts a familiar privacy problem in a different place. A phone is usually visible when someone holds it up to take a picture, while glasses sit in the same position whether or not their camera is in use.

The company says the eyewear can use what a wearer sees to help with activities such as reading a menu or translating a sign. Google also describes an accessibility use in which a spoken description can help a person who is blind or has low vision identify an item or read small print. Those uses depend on access to a view of the wearer’s surroundings.

What the wearer can control

Google says camera visuals used by Gemini are not stored in Gemini History, used to train AI models or used for ad targeting. It distinguishes camera data from the conversation history settings available to a wearer. The company also points to app permissions and a physical power switch as controls that do not rely solely on a person remembering to end a conversation.

These are statements about Google’s design. They should not be read as a promise that all information connected with every use of the glasses disappears, or that a photo a wearer chooses to capture cannot be retained or shared. Buyers evaluating the product will need to consider how the rules apply to the specific feature they use, including the difference between getting an AI response and recording media.

What others can see

Google says a light facing outward signals when the wearer takes a photo or records video. The company also describes automatic face redaction and directional microphones as protections for people nearby. The real-world usefulness of these protections will depend in part on whether people can recognize an indicator and understand what it means, including at an angle or in bright light.

The privacy considerations extend beyond recording. A camera worn at eye level can be pointed at documents, computer displays or people in workplaces where phones and cameras are already restricted. Organizations that allow wearable devices will need to decide whether existing rules cover AI features as well as conventional photos and video.

Design claims still need practical checks

Google’s announcement explains how it intends to reduce risks for both wearers and bystanders. It does not establish how reliably face redaction will work in every setting, or whether a nearby person can distinguish all the ways the glasses use a camera. Those are useful questions for product testing. The distinction is especially important when people near the wearer cannot see the assistant’s interface or review its privacy settings themselves.

Stay on top of AI

Get the stories that matter in artificial intelligence, straight to your inbox.

Free. One email with the stories that matter. Unsubscribe any time. See our privacy policy.