How do notice, consent and data rights work when the people being processed are not the people using a device? Smart glasses can capture faces, voices and surroundings while the wearer simply looks at someone. AI pushes the issue further because that data collected can be analyzed, transmitted or reused.
European regulators are starting to treat this scenario as its own category of privacy risk (Cybernews, 2026). The European Data Protection Board agreed in June to prepare public recommendations on smart glasses by the autumn, after several national regulators raised concerns about their growing use. Hamburg’s privacy authority has been examining Meta’s Ray-Ban glasses since 2025, including the visibility of their recording indicator, the data transmitted by the device and how information may be used for AI training.
Meta’s main signal to people nearby is a white capture LED. It flashes while photos or video are being taken, and newer models disable the camera if the light is blocked or physically tampered with.
Reasonable engineering, perhaps, but socially it asks a lot in very low light. The person being filmed has to notice it, understand what it means and object before the recording goes anywhere else. Such indicators can be ineffective depending on distance, lighting and whether people around the wearer even recognize the signal. Unlike a phone being raised and pointed, the glasses themselves provide very little behavioral warning that recording has started.
Hamburg’s regulator has been testing whether the LED is actually visible enough in real-world situations, while commissioner Thomas Fuchs has gone further, telling German media that smart glasses could fall foul of national rules against recording devices concealed inside everyday objects (Euractiv, 2026). He had even raised the possibility of a ban, although Germany’s telecommunications regulator holds the final say on legality. France’s CNIL is already warning that wearable cameras could turn familiar, fixed surveillance into something mobile and much harder to recognize. In a survey of 2,128 French adults, 67% said smart glasses posed a privacy risk.
The privacy interface has effectively moved from an obvious human gesture to a hardware notification on somebody else’s face.
Recording is only the first layer. Smart glasses increasingly use cameras and microphones as inputs for AI, allowing someone to ask the device about the person, object or conversation in front of them. CNIL notes that Meta’s glasses are closely tied to its AI ecosystem, with voice, image and video inputs sent to company servers when needed to generate responses. It also raises older privacy problems behind that processing, including reuse, international transfers, subcontractors, and human annotation.
For bystanders, the arrangement is oddly one-sided. The buyer gets an account, privacy settings and terms of service. The person sitting opposite them in a café gets none of those things, yet their face, voice and surroundings can still become input. And AI could make first-person data considerably more desirable. CNIL’s research arm has already identified smart glasses as a potential source of training data for world models, systems designed to learn from video, audio and other sensory streams rather than mainly text. First-person footage is especially attractive because it captures movement and physical interaction from roughly the perspective of a human observer.
So, a mundane interaction can acquire a second life as data without the other person ever knowing it happened.
Smart glasses bring several problems that used to live in separate products onto one face. The camera resembles a body-worn surveillance device and the microphone behaves like a voice assistant. AI can interpret the resulting stream, while cloud infrastructure determines where it travels and who may process it.
That is a harder privacy problem than deciding whether a camera light is bright enough.