Meta AI glasses are face-worn cameras, microphones, speakers, sensors, and cloud AI tied to a Meta account. [wrapped in something that looks like normal eyewear]
That is the threat model.
When someone records with a phone, there is a visible gesture. People notice the rectangle.
With AI glasses, the camera sits at eye level. The social signal is gone. Bystanders may not know they are being filmed, recorded, transcribed, translated, or fed as visual context into a Meta AI query.
The capture LED is NOT consent. It is not visible in every environment. It does not solve audio. It does not solve bystander privacy. It does not solve the fact that most people will not understand what the device is doing.
Meta tells users to power the glasses off in doctor’s offices, bathrooms, locker rooms, schools, and places of worship. That advice is the tell. If a product needs users to remember to disable it in sensitive environments, the default posture is already broken.
The data flow: Calendar, Outlook, Garmin, Strava, Apple Health, Health Connect, WhatsApp, Messenger, Instagram, voice commands, cloud inference, all routed through Meta. Your calendar alone is a social graph, location history, medical appointment log, work map, and routine fingerprint. Now strap that context to a sensing platform on your face.
The Ray-Ban Display model adds in-lens notifications, live captions, navigation, AI answers, and wrist neural input. The attack surface expands. Shoulder surfing the lens. Visual prompt injection from the environment. Malicious app connections. Account compromise exposing message and device controls.