The Electronic Frontier Foundation has warned that a growing category of AI-enabled glasses, watches, pendants and other wearables is changing privacy from an individual product choice into a decision imposed on everyone nearby. The advocacy group says technical protections for device owners do little for bystanders whose speech or image may be captured without a practical way to refuse.
EFF points to camera-equipped smart glasses from Meta and Google, Apple's Audio Intelligence feature and Amazon's Bee devices as examples of products that can record, transcribe, describe or summarize a user's surroundings. Many other companies are exploring devices intended to remain active throughout the day. Their appeal is low-friction assistance, but that same quality makes data collection less visible than deliberately raising a phone or recorder.
Companies commonly emphasize controls available to the customer: local processing, deletion of raw audio, configurable privacy settings, face blurring or terms that prohibit recording without permission. EFF argues that these safeguards focus on the purchaser rather than the person standing beside them. Even when raw audio is discarded after transcription, a microphone still captures a conversation long enough to turn it into text or a summary.
The central problem is consent in shared physical spaces. A colleague, friend, patient or stranger may not recognize every device or understand whether it is active. Indicator lights and phone alerts can disclose that recording is occurring, but notification alone does not give a bystander a realistic choice, particularly at work, on public transport or in other settings where leaving carries a cost. Rules also vary across jurisdictions and social contexts.
EFF further warns that transcripts and summaries may become targets for legal demands. Products differ in how data is stored and how difficult outside access may be, so the existence of derived records can matter even when companies say they do not retain original audio. The group calls for closer scrutiny of those technical differences rather than treating all ambient assistants as equivalent.
Beyond data security, the organization identifies a social effect. People behave differently when they believe ordinary conversations may become permanent, searchable records. Continuous capture can reduce spontaneity and collapse distinctions between situations where note-taking is expected and intimate conversations where it is not. Devices cannot reliably interpret those unwritten boundaries.
The critique does not establish that every announced product will gain broad adoption. EFF notes that technology categories often disappear when consumers reject them. That uncertainty makes public expectations and purchasing choices relevant, alongside regulation and engineering.
For manufacturers, the challenge is broader than protecting an account holder's information. Meaningful privacy design would need to address notice, consent and control for non-users who never agreed to the device's terms. EFF's position is that transferring that responsibility to individual owners is inadequate when the product is designed to fade into the background.
