Privacy audit reveals mandatory cloud voice storage and AI training risks in Ray-Ban Meta glasses
Updated terms quietly eliminate voice storage opt-outs for US users, requiring manual deletion while routing multimodal vision queries into training pipelines.
Technology news
A comprehensive privacy evaluation of the software ecosystem powering Ray-Ban Meta smart glasses has revealed significant telemetry and data retention shifts that quietly remove key consumer opt-outs for United States users.
Under Meta’s updated service policies, US wearers are no longer permitted to toggle off the cloud storage of audio recordings generated during voice interactions. While the companion Meta View application allows users to periodically review and manually purge historical voice logs, all queries initiated by the 'Hey Meta' wake phrase are automatically transmitted and archived on remote servers by default.
The audit also highlights substantial privacy implications surrounding the glasses' multimodal AI capabilities. When wearers ask Meta AI to analyze their surroundings—such as identifying monuments, translating printed text, or evaluating products—captured imagery is uploaded to Meta's infrastructure and processed for AI foundation model training unless users navigate through multi-layered settings to disable optional data sharing.
Because smart glasses record both wearers and non-consenting bystanders in public spaces, privacy researchers urge users to take active precautions: disabling always-on wake phrase listening to prevent accidental trigger events, opting out of model-training data ingestion, and severing automated cloud photo backups to prevent passive biometric harvesting.