Know what’s on your site before a plaintiff’s firm does
A free privacy audit shows which cookies, pixels and trackers are running — and which ones commonly appear in privacy claims.
The Netherlands’ data protection authority, the Autoriteit Persoonsgegevens (AP), has told consumers not to buy AI camera glasses for now. After testing six models sold on the Dutch market, the regulator concluded on November 19, 2026, that it “considers AI camera glasses unsafe at this time due to the risks to users and bystanders.” The findings go beyond the familiar worry about covert filming. In simulated emergencies, one pair of glasses claimed an ambulance was on its way when nobody had been called. For manufacturers, app developers and the businesses whose premises these glasses walk into, the report is a detailed list of what a regulator now expects.
TL;DR
- The Dutch AP tested six AI camera glasses in September and October 2026 and advised consumers, in general terms, not to buy one at this time.
- It found recording lights that were hard to see and easy to tape over, cloud AI processing often outside the EU without clear disclosure, serious security flaws and dangerous AI advice in safety scenarios.
- The AP did not call for a ban, but asked manufacturers to build in safety, privacy and AI Act compliance and advised venue owners to prohibit camera glasses in their house rules.

What the AP tested, and in what capacity
The study was published as a special edition of the AP’s twice-yearly AI and algorithms risk report, titled “AI-camerabrillen onder de loep” (AI camera glasses under the microscope) and dated November 18, 2026. Two details shape how it should be read.
- It is not a GDPR investigation. The report says the AP acted as the Netherlands’ coordinating supervisor for algorithms and AI and tested product safety “from an AI perspective.” No company was found to have broken the law, and no fines were issued.
- The products are anonymous. The AP selected six glasses at random across price ranges. It has declined to name them, telling Dutch and British media only that they come from different continents and cost from under €100 to over €500.
Two of the six were put through detailed scenarios in which user safety mattered. The AP also examined design, documentation, data flows and security. A second report, on audio-recording and transcription wearables, is planned for 2027.
What the tests found
The AP’s announcement and report group the problems into four areas.
Bystanders can barely tell they are being filmed
Only one of the tested glasses had a recording light visible enough, from a distance and in everyday conditions, to protect people nearby. Many lights were white, which washes out in daylight and differs from the red light people associate with a running camera. Some were dim or did not blink. The lights could be covered with tape with almost no effort, and all but one pair kept filming when they were, even though manufacturers said that would not happen. One manufacturer even advertised how easy it was to cover the light. During testing, one pair also made a recording without the wearer noticing.
Data goes to unclear destinations
The AI features ran entirely in the cloud: the glasses send images and audio through a smartphone app to external servers for processing. Only a few products told users that an AI model was processing their recordings, and privacy statements said little or nothing about which models were used or whether recordings trained them. Images and text from conversations with the assistant went to servers outside the EU without proper notice. In one case, the AP saw traffic flowing to a server in China while the app’s privacy statement named Singapore.
Security flaws let strangers take over
In several glasses the AP found a serious vulnerability: someone nearby could connect without the wearer’s credentials, take control, capture photos and videos and pull recordings not yet transferred to the phone. Some back-end servers showed signs of running outdated software that had not received security updates for a long time.
The AI gave dangerous advice
In a simulated resuscitation, one pair repeatedly said an ambulance was on the way although no emergency services had been contacted. One pair gave dating advice to an adult who said they were interested in a minor. The glasses also failed to follow safety warnings on toys for young children and made errors reading medicine leaflets and severe-weather warnings.
Compliance paperwork was weak too. Several products did not clearly identify the manufacturer or an EU representative, none included a statement of conformity with the EU AI Act, and a few carried a misleading mark that closely resembled the CE marking.
AP vice chair Monique Verdier summed it up: “A pair of glasses that films you unnoticed is already worrying. A pair that also gives you the wrong advice when someone’s life is at stake goes a step too far.”
Where the household exemption stops
Much of the debate about smart glasses turns on a single GDPR clause. Article 2(2)(c) says the regulation does not apply to processing “by a natural person in the course of a purely personal or household activity.” A tourist filming a canal for a holiday album is usually outside the GDPR.
That exemption is narrower than many wearers assume, for three reasons:
- Sharing ends it. In its September 22, 2026 guidance for wearers, the AP said the exemption may cover strictly personal use, but no longer applies once footage of recognizable people is published or shared more widely. It added that in public spaces, informing everyone and getting consent is often practically impossible, which is why it says filming others is “almost always” prohibited.
- Public space is not a home. The Court of Justice of the EU held in the Ryneš case (C-212/13) that a home camera which also covered part of a public space was not a purely household activity. The reasoning maps closely onto a camera worn on someone’s face in the street.
- Providers are never exempt. Recital 18 of the GDPR states that the regulation applies to controllers or processors that provide the means for personal or household processing. A glasses maker or AI provider that receives bystanders’ images on its servers is processing their data in its own right, whatever the wearer’s status.
The AP’s September guidance also made the AI point explicit: even when no photo or video is saved, asking an assistant about what the camera sees can involve processing the personal data of the people in view, people who, as board member Katja Mur put it, “did not choose this themselves.”
What makers, app providers and venues should do
The AP’s call to manufacturers is short: bystanders must be able to see clearly when they are filmed, the light must not be easy to defeat while recording continues, users must know where their data goes and who processes it, AI must be reliable when health or safety is involved, and products must be secure and demonstrably compliant with EU rules including the AI Act. Turned into a working checklist:
- Make recording impossible to hide. Use a bright, blinking indicator and stop capture if the light is covered. The AP notes this is technically achievable.
- Treat bystander data as your own processing. Run a data protection impact assessment covering people in view, decide a lawful basis for any server-side processing, and minimize what is retained.
- Disclose the AI supply chain. Name the model provider, the server locations and the transfer mechanism, and say plainly whether recordings train models. Make sure the privacy statement matches actual traffic.
- Close local attack paths. Require authentication for any device pairing, patch back-end servers and test for nearby takeover before shipping.
- Guard safety-critical answers. Block false claims about calling emergency services, add protections for minors and test medical and weather scenarios.
- Get the EU paperwork right. Show the manufacturer or EU representative, a genuine CE marking and documentation of AI Act obligations.
For organizations on the receiving end, the AP advises owners and operators of locations to ban wearing and using camera glasses in their house rules, and it has published a standard Dutch-language clause to copy. That matters especially in hospitals, courts and places of worship, which the AP singled out. Employers face a parallel question, since staff may already bring these devices to work; our guide on employees wearing cameras without a policy covers how to address it.
The Dutch study adds to growing scrutiny elsewhere. Australia’s privacy commissioner has flagged risks from consumer smart glasses, Meta’s glasses have faced a high-profile lawsuit, and reporting showed Meta removed a hidden facial recognition system from a companion app. For a broader primer on the category, see our overview of wearables and data protection risks.
Frequently asked questions
Has the Netherlands banned AI camera glasses?
No. The AP advised consumers not to buy them for general use at this time and asked Dutch and European policymakers to weigh its findings. Verdier said simply banning everything is not sustainable in the long run.
Which brands did the AP test?
The AP has not said. It describes six models available in the Netherlands, from different continents and price points, and says every pair had shortcomings.
Is it legal to wear camera glasses in public in the EU?
Wearing them is not prohibited by the GDPR, but filming recognizable people and sharing or publishing the footage generally requires a legal basis, and the household exemption usually will not cover it. Venues can also set their own rules.
Does the AP see any legitimate uses?
Yes. It notes the glasses can help people with visual impairments or low literacy understand their surroundings, and says those users in particular need technology they can trust.
Look before you wear
The AP’s message is not that seeing-and-talking glasses have no future. It is that the products on shelves today ask bystanders to trust a tiny white light, and ask wearers to trust an assistant that can invent an ambulance. Until makers fix both, the regulator’s advice is to keep them in the box.
Put this into practice
Turn privacy guidance into working controls.
Captain Compliance helps teams discover tracking technologies, enforce consent, manage privacy requests and document the evidence behind every decision.