How to Stop Meta Using Your Smart Glasses Visual Data for AI Training
Meta has added an opt-out that stops visual data from its smart glasses being used for AI training. The switch is in the Meta AI app and applies to data collected during AI experiences on the glasses Engadget.
Meta is showing owners a pop-up in the app. The notice says that if you opt out, that data is not kept after it has been processed. To change it, open settings for your paired glasses, choose "Meta AI" and scroll to "store visual data from AI experiences."
The setting applies to each pair of glasses separately. The glasses must be out of their case and connected to your phone to change it. If you own more than one pair, you must repeat the step for each pair. Storage stays on unless you turn it off on every unit.
Meta says it saves the resulting visual data when you use AI experiences. That data can be used to run glasses features and to improve Meta products, including training AI models through automated systems or manual human review. The company says it only saves visual data from certain AI features, including Live AI, the mode that can see what is in front of you and respond in real time. It says visual data does not include photos and videos you take when the capture light is on.
Turning off storage has a tradeoff. Meta warns that some features may stop working, including "visual reminders." That is the feature where you look at an object and ask the assistant to remind you about it later. That reminder needs a saved image to work.
For translation, an earlier explainer showed how opt-out works in practice. If you have opted out and ask the glasses to translate a menu, Meta uses the image to produce the answer and then deletes it Engadget. The instant answer, known as inference, still happens. Saving the data to improve products does not. Meta's help pages say that when opted out, visual data is not used for product improvement.
The broader context here is how long wearable cameras should keep what they see. A UK pub chain has restricted Ray-Ban Meta glasses over concerns about secret recording, and Meta faces a privacy lawsuit connected to its smart glasses Reuters. Meta has also built camera-free smart glasses meant to reduce criticism of devices that record video and photos Reuters. Earlier objections in the EU focused on plans to use personal data to train AI models without asking for consent Reuters.
Looking at what this means for deployment, the system keeps two jobs separate: answering in the moment, and saving data to build future models. Business buyers, developers building on Live AI, and privacy reviewers should treat the switch as a control on each device, not as an account-wide guarantee. Human review is still part of the process for data that is kept, which leaves open questions about blurring sensitive details, who can view images, and how first-person photos are tracked.
In my view, the opt-out works but puts the effort in the wrong place at scale. A per-device opt-out is easy to miss across households and large fleets. The positive side is real. Processing a frame to answer and then discarding it shows useful glasses do not have to add images to a training set. If Meta can keep that no-storage option working as well for core tasks, privacy-friendly defaults become easier to support without breaking features.
