Google said in a post on 12 November that its Gemini glasses are designed to remove bystanders' faces before Gemini processes an image. It also said these images are not saved in Gemini History or used to train AI. When a wearer takes a photo or video, an outward light comes on. Google said that if Gemini is asked to take one, it is stored in the wearer's gallery.
MediaNama reported that Google describes face removal only for images sent to Gemini. Google's wording suggests the step happens on the glasses, before the image is sent. The post does not say whether the same step applies to gallery captures, or to what apps with camera access receive. For recordings, Google points to the indicator light, sensors that block capture if the light is covered, and usage rules. MediaNama sent Google questions on 13 November and had no reply when it published.
The protection is tied to one data path, not to the camera. An image sent to a model for one answer is different from a photo or video kept as a file in a gallery. Organisations weighing these glasses can ask Google in writing whether face removal covers gallery captures and what permitted apps receive.