Four taps in the Meta AI app now stop Meta from keeping what your glasses see when you ask its AI about it. The switch is called “Store visual data from AI experiences”, and it began reaching owners in the week of September 23, 2026. With storage off, Meta says the images and video behind a live AI session or a “what am I looking at” question are used to answer you. After that they are not stored, not used to train its models and not shown to human reviewers. Meta’s help page lists what the switch leaves out, including your voice, the text descriptions the AI writes and anything processed before you flip it.
Switching it off
Open the Meta AI app, go to your glasses’ settings, tap Meta AI under Experiences and turn off Store visual data from AI experiences. Meta’s help page says you can change the preference again at any time in the same menu.
The same choice appears during setup and in an in-app notice. The help page does not say whether the setting starts on or off for existing owners, so if you have never seen the notice, open the menu and look.
What counts as visual data
Meta doesn’t count the photos and videos you take on purpose, with the capture LED on, as visual data. Those stay on the glasses until you import them to your phone.
Visual data is what the camera sees for the AI: the frames sent when you ask about a sign, translate a menu or keep live AI running while you cook. With storage on, Meta may use that material to improve its products, including training that “may be automated or manual (human)”. Those frames often include people around you who never agreed to be filmed, which is the concern the Dutch privacy regulator raised about camera glasses on September 22.
The five things the switch does not cover
Meta’s help page lists them:
- Voice interactions with Meta AI
- Images and videos you share with Meta AI elsewhere, such as in Messenger or Instagram
- Images and videos that may break Meta’s terms, for example possibly illegal content
- Metadata linked to visual data, which Meta says includes AI-generated text descriptions of an image and camera exposure settings
- Visual data processed before you turned storage off
When you ask “what am I looking at”, the AI turns the picture into words before it answers. Meta’s help page counts that description as metadata, not visual data, so turning storage off doesn’t stop Meta from keeping it. The page doesn’t say how long those descriptions are stored or whether people review them.
The last item means the switch works forward only. Whatever was processed before you turned it off stays wherever Meta already put it.
Your voice falls outside the switch
Voice interactions top Meta’s list, so turning off visual storage doesn’t change anything Meta keeps from what you say. Check the voice settings in the Meta AI app separately.
What switching it off costs you
The help page says some features may be limited or unavailable without stored visual data, and the one it names is visual reminders. The case for leaving storage on is an assistant that remembers your preferences over time. Owners who ask their glasses where they parked will feel the loss, but a question about a street sign gets the same answer either way.
The promise Meta made the same day
On September 23, Meta’s engineering blog also described Private Processing for its AI glasses. With it, requests run inside sealed servers where, Meta writes, “even Meta cannot access your data”. The post gave no date, no list of glasses and no word on whether the feature will be on by default.
