Meta lets users opt out of AI training on smart glasses visuals

Meta allows users to opt out of having their visual data from its smart glasses used for AI training.

Opt-out covers camera-based AI interactions. Users can prevent images and video from Meta AI sessions on the glasses from being used for model training. Those images will also no longer be sent to third-party contractors for review.

The change applies to any camera interaction with Meta AI, such as asking the glasses to translate a menu or explain a sign. Meta stated that once a user opts out, the visual data is used only to answer the immediate query and then deleted.

Previous practice stored unsaved captures. Meta had already said that photos saved to the camera roll are not used for training. However, images captured during AI interactions were stored and sometimes reviewed even though they do not appear in the user's camera roll.

Those recordings were occasionally sent to contractors in other countries who labeled the images for AI training. Some contractors reported seeing personal and intimate images during this process, according to an earlier report.

Audio data stays available for training. The opt-out applies only to visual data. Audio interactions with Meta AI can still be used for training and review, though users retain the ability to delete those recordings manually.

The opt-out addresses concerns raised by privacy critics and at least one proposed class-action lawsuit. Meta also revealed a new lineup of Ray-Ban-branded glasses that contain no cameras at all.