Meta is addressing privacy concerns by letting users opt out of having visual data from their smart glasses used for AI training. The new setting applies to camera-based interactions with Meta AI.
Previously, photos and videos captured during AI interactions could be used to train company models. These captures are different from standard photos and are not saved directly to a user phone camera roll.
Meta acknowledged the confusion surrounding these captures. Any camera interaction with the AI, such as translating a menu or asking about a sign, stores the resulting images.
Some of these recordings were previously sent to third-party contractors in other countries. Those contractors reviewed and labeled the images for artificial intelligence training purposes.
A report earlier this year revealed that contractors occasionally saw personal and intimate images during this review process. The new privacy setting specifically addresses this visual data collection.
Meta explained how the system works for users who choose the opt-out option. If someone asks the glasses to translate a menu, the visual data is used to answer the question and is then deleted.
The company confirmed that the data is not used for training. Nobody will review it for product improvement purposes under the opt-out setting.
The new control only covers visual data captured by the devices. Audio interactions with Meta AI remain usable by the company for training and review.
Users can still manually delete their audio recordings. The opt-out update helps Meta respond to privacy critics and a proposed class-action lawsuit.
Meta also revealed a new lineup of Ray-Ban-branded glasses that do not include cameras at all.
