Key facts
- The authority found recording indicators could be covered without stopping the camera.
- AI advice was unreliable in simulated safety scenarios.
- The regulator advised consumers not to buy the devices for general use.
- Two models underwent deeper testing in mock personal-safety scenarios.
The Dutch Data Protection Authority tested six AI-powered camera glasses sold in the Netherlands and found their privacy and security safeguards fail during testing.
Biometric Update reported the authority’s findings. The regulator said recording indicator lights could be covered without stopping the camera, and most tested devices kept recording after the light was obscured. On virtually all devices, the lights were hard to see or easy to cover. That leaves people facing a wearer with few dependable ways to tell if recording is happening.
The authority also found security weaknesses that could let someone nearby connect without permission, access recordings, or operate the camera. Several devices had these weaknesses. Users were not always clearly informed about where images and audio were processed. They also were not always told which companies participated or which AI systems obtained the material.
AI function data was frequently processed through servers located outside the European Union. The authority uncovered another category of problems when testing the glasses as AI assistants. Two models underwent deeper testing in mock personal-safety scenarios.
In one simulated emergency resuscitation request, a device repeatedly told the wearer that help was already on the way, although no emergency services had been contacted. Devices also made errors concerning drug information, severe weather alerts, and toy safety. In another test, a device gave dating advice to an adult who expressed interest in a minor.
The regulator said these failures are especially dangerous because AI-generated responses can sound authoritative despite being incorrect. It advised consumers not to buy the devices for general use. The six models available in the Netherlands were priced from below €100 to more than €500.
The regulator acknowledged AI glasses may offer significant benefits to blind or visually impaired people and to people with low literacy, recognizing their value for accessibility. But it argued those uses make reliability and security more important rather than less.
The authority declined to name the manufacturers or AI models because the research was a sample rather than a comprehensive product ranking. It also said it did not identify Meta glasses as among the tested products, so its findings cannot be attributed to Meta or another manufacturer.
The findings come weeks after the European Data Protection Board issued guidance calling for stronger recording indicators. In October, the board said indicators must work across lighting conditions and catch the attention of anyone in view of the camera. The board said indicators should remain active long enough for people to understand images or video are being captured, including for AI assistant processing. Recording information should not be possible when an indicator is covered, and transparency tools built into technology must actually work and be hard to bypass.
The Dutch results test an industry privacy defense against an emerging European regulatory standard and show at least some devices do not meet it. A September Biometric Update report said Meta’s glasses can power camera-based AI functions without turning on the same indicator light used for photos and video. Meta explains the camera image serves an AI query instead of being stored in the wearer’s gallery. That creates a gap between a recording warning light and AI tools that can examine what a wearer sees without the same visible warning.
Under the GDPR, the personal or household activities exemption does not automatically cover manufacturers or AI providers that process the information. According to the European Data Protection Board, a smart-glasses maker or provider of a linked app or AI service can be a data controller if it decides how personal information is processed. The board points to an AI provider examining a camera image that includes other people. Such providers cannot rely on the wearer to meet obligations to people whose personal data is captured.
The warning arrives amid a broader European backlash as authorities and businesses confront eyewear that puts cameras, microphones, AI, and biometric identification into frames made to look like regular glasses. The pressure is especially strong in the Netherlands. Consumentenbond, a consumer organization, has sued Meta over claims its camera glasses breach Dutch and European privacy law by filming non-consenting bystanders. A hearing has been scheduled for November 26 in Amsterdam.
The organization wants changes to the product and its settings and says Meta should stop sales if it cannot make the glasses compliant. In October, the Dutch eyewear chain Hans Anders stopped selling Ray-Ban Meta glasses because of privacy concerns in the Netherlands and Belgium.

Join the discussion