The European Data Protection Board's preliminary guidance on smart glasses

On 13 October the European Data Protection Board adopted a statement on smart glasses, noting that because they look much like ordinary glasses, people nearby may be filmed or recorded without knowing it. The statement recommends that manufacturers signal recording with reliable lights and sounds that are hard to switch off, and that recording should not be possible while the light is covered. The press release presents the statement as preliminary guidance, and the statement sets no new rules and names no manufacturer.

On 13 October the European Data Protection Board (EDPB, the body made up of the data protection authorities of the EU member states) adopted a statement on smart glasses, setting out under the EU's General Data Protection Regulation (GDPR) the responsibilities of users, manufacturers and service providers in the EU. According to the statement, smart glasses bring new uses for communication and assistance, such as real-time translation and support for people with visual impairments.

According to the statement, these glasses look much like ordinary glasses, so collecting and recording personal data is barely noticeable. Microphones on some models can pick up background conversations the wearer is not part of. The statement adds that combining smart glasses with face or voice recognition systems poses a high risk to rights and freedoms in public spaces.

Users who record only for a purely personal or household activity, such as sharing only with family or friends, are not subject to the GDPR. The statement says this exemption must be narrowly construed, and other laws, such as criminal law, may still apply. Users who share to a public social media account or use the glasses at work become controllers (the party that decides the purpose and means of processing). Manufacturers and providers of AI features are also controllers where they decide the purposes and means, and the statement says they should not rely on users to meet their duty to inform people nearby.

For manufacturers, the statement recommends combining several signals. A light should be visible in many lighting conditions, recording should not be possible while the light is covered, and a shutter sound can be added. Phone notifications that detect nearby glasses can help but should not be the only signal, and should not allow the wearer to be tracked. The EDPB's press release presents the statement as preliminary guidance; the statement itself sets no new rules, names no manufacturer and closes by saying the EDPB will continue to analyse the privacy concerns smart glasses raise.

Perspective

The light is the main cue people nearby have that they are being filmed. We could not reach Meta's own website, so Meta's statements here come from its newsroom post as quoted by Privacy Guides and 9to5Google. Every pair has a white capture LED on the front that blinks briefly for a photo and keeps blinking during video, and the LED has no off switch. From the second generation onwards, the camera is disabled when the glasses detect that the LED is blocked, and a July update also disables the camera if the LED has been tampered with or destroyed.

According to PPC Land's paraphrase, Meta considers a shutter sound impractical for notifying people at a distance and points to precedents such as laptop camera indicators for its use of a light.

In September the data protection authority of Hamburg in Germany published an inspection report on Ray-Ban Meta, and as netzpolitik.org reports it, the report's conclusion is that the LED is only of limited visibility depending on distance, angle and lighting. CORRECTIV reports two further findings. Blocking is checked only when recording starts, so covering the light mid-recording does not stop it, and on the second generation the light stays off entirely while AI features are in use (netzpolitik.org only mentions an inspection of the first generation). A footnote in the EDPB statement adds that a white LED could be mistaken for a reflection.

Section 8 of Germany's Telecommunications Digital Services Data Protection Act (TDDDG) bans devices disguised as everyday objects and designed for covert audio or video recording, and the Federal Network Agency (Bundesnetzagentur) is the responsible authority. Asked by CORRECTIV, the agency assessed the LED as clearly recognisable within 5 metres and found the conditions for a ban not currently met. The agency also said that the mere possibility of modification is no reason for a ban, and that it had not yet examined the second generation.

TNW reported that a Meta executive announced on Threads on 27 August an update that stops the camera when the light is covered, and described it as closing the gap of starting a recording and then covering the light. That is TNW's account. The report does not say which firmware version Hamburg inspected, and we cannot confirm that the two refer to the same gap.

In October, users reported that Ray-Ban Meta glasses had saved recordings of conversations without the wake word being spoken. Meta's response to Android Authority, as quoted by Privacy Guides, says the issue has been resolved and that users can manage, review and delete their voice recordings in settings at any time. As of 29 October, none of the reports we read gave the cause or the number of people affected. The reports we read also mention only voice recordings, not photos or videos.

Ray-Ban Meta wearers concerned about accidental recording can review and delete voice recordings in the glasses' privacy settings in the Meta AI app. According to TechRadar, deletion only works after the fact and cannot prevent capture. TechRadar also writes that turning off voice activation seems to stop the problem, while ChannelNews notes that the cost is losing hands-free access to the assistant, which can still be called up with the button.

France's data protection authority, CNIL, also has recommendations for wearers. CNIL advises telling people nearby, switching off capture features when not needed and turning off the glasses wherever people are asked to turn off their phones, and suggests getting consent before posting images that show other people on social media.

People being filmed have fewer options. CNIL's page says lights that inform people nearby have only limited effect and are absent for some uses, and the EDPB statement says phone apps that detect smart glasses can only tell that glasses are nearby, not whether they are recording. If it matters to you, the simplest step is to ask the wearer directly whether they are recording (our suggestion).

Singapore's Ministry of Digital Development and Information (MDDI) answered two parliamentary questions on smart glasses in 2026. In its 8 April answer, the ministry wrote that smart glasses with AI capabilities "are subject to existing safeguards for data protection and safety in Singapore". Organisations offering them must comply with the Personal Data Protection Act (PDPA), and an organisation that transfers such data to third-party contractors overseas remains responsible for protection comparable to Singapore's standards.

Asked on 6 May whether it would introduce visual indicator requirements, the ministry replied that photography and recording are "subject to the same rules and principles regardless of the form factor of the device". According to the answer, organisations, including individuals creating content for commercial purposes, are accountable under the PDPA, which generally does not require them to seek consent for photography or recording in public spaces, but they must ensure that any later use or disclosure of the data, including for AI features, is for a reasonable purpose. The answer announced no indicator requirement. Both answers state that non-consensual recording, sharing or other misuse of intimate or private content may be a criminal offence under the Penal Code and the Protection from Harassment Act.

In South Korea, the news agency Newsis reported on 3 August that, according to the Personal Information Protection Commission (PIPC), the commission met Samsung staff in late July about the AI glasses Samsung plans to release within the year. They discussed notification devices, such as a light or sound, that make it easy for people nearby to notice when the camera and microphone are in use. The PIPC had earlier met Meta, which has launched its AI glasses in South Korea, and the report says the government plans to draw up privacy principles for AI glasses and autonomous robots with industry in the second half of 2026. We did not obtain a PIPC press release, so this account rests on the Newsis report.

As of 29 October, we found no statement or guidance specifically on smart glasses from Japan's Personal Information Protection Commission (PPC). A compendium of the approaches of G7 data protection authorities, compiled by CNIL in 2026, summarises earlier PPC recommendations on camera systems with face identification, and the compendium lists the PPC's view that camera images can fall within the scope of the Act on the Protection of Personal Information if they can identify a specific individual.

We, a community based in Taiwan, found no rules in Taiwan specific to smart glasses and no explanation from the authorities as of 29 October. Article 315-1 of Taiwan's Criminal Code punishes recording another person's non-public activities, speech or conversations without lawful reason, where non-public means not open to others, so that people do not expect to be heard or seen. We did not find the original text of any court rulings on whether filming in a public place counts as non-public.

Article 51 of the Personal Data Protection Act of Taiwan excludes purely personal or household activities, as well as audio-visual material collected in public places or at public events that is not combined with other personal data. Going by the wording, our reading is that footage combined with other personal data, for example when AI is used to identify who someone is, falls outside that exclusion.

Readers can compare approaches with a few questions. One is whether, when people nearby are filmed, the law looks only at whether the place is public or also at how the footage is used afterwards, as Singapore's answer does for organisations. Hamburg and the Federal Network Agency reached different conclusions on how visible the light is, after examining different models in different situations, so it is also worth comparing who should carry out such tests and against what standard, and whether indicator requirements belong in law or in agreements with industry. It is also worth comparing how each place's rules preserve features that help users, such as assistance for people with visual impairments and real-time translation.

Get new stories by RSS, newsletter or Bluesky