The Dutch Data Protection Authority declared AI smart glasses unsafe after tests revealed faulty emergency advice, security flaws and obscured recording lights.
The Dutch Data Protection Authority has warned consumers against buying AI-powered smart glasses with cameras, declaring the devices unsafe for wearers and bystanders this week. In test findings, the regulator reported that the gadgets carry serious hazards, including bad advice during simulated emergencies and cybersecurity vulnerabilities that allow strangers to hijack device cameras.
Faulty advice and safety hazards
During testing across six models, the Dutch regulator, known as the Autoriteit Persoonsgegevens or AP, observed critical errors when the glasses interpreted real-world tasks. In a simulated cardiopulmonary resuscitation emergency, one pair told the user an ambulance was on its way even though emergency services hadn’t been called, TechRadar reported. In another trial, a pair offered dating guidance to an adult who asked about pursuing a minor.
The agency also documented devices failing to process warnings on children’s toys, misreading medication leaflets, and misinterpreting severe weather alerts, The Register reported. The AP tested glasses priced between under €100 ($110) and over €500 ($550). Tested brands that sell in the Netherlands included Meta, China’s TCL/RayNeo and Rokid, and Taiwan’s HTC, though officials declined to link specific failures to individual products or price points.
“Glasses that film you unnoticed are already concerning. Glasses that also give you incorrect advice when someone’s life is at stake go a step too far,” AP vice chair Monique Verdier said.
Security flaws and hidden recording
Security assessments uncovered technical vulnerabilities alongside safety blunders. Regulators discovered that nearby individuals could connect to some glasses without permission, inspect stored recordings, and take over the camera.
The watchdog found physical recording indicators were dim and easily blocked on almost every tested model. Testers could also manipulate devices to record footage even with the light covered, contradicting marketing promises from hardware makers. Unlike smartphones, which require overt body language like holding a screen up, smart glasses resemble normal eyewear and lack visible signals during recording. This stealthy form factor has fueled public pushback, including online mockery labeling them “pervert glasses” or “perv glasses” and a “Ban Ray” boycott encouraging shoppers to stick with regular sunglasses.
Data handling raised further regulatory problems under European rules. The AP said AI features rely on processing personal data on servers mostly based outside the European Union, while wearers receive inadequate explanations about how their captured information is processed.
Bans and upcoming policy
The watchdog acknowledged helpful uses for the technology, including assistance for people with lower literacy or visual impairments. Still, the AP maintained its position that current camera glasses remain unsafe for users and the public.
The advisory follows earlier moves to curb the devices. In September, the AP reminded the public of personal privacy rights and urged venues to bar smart glasses. Other organizations have already imposed restrictions. The DEF CON hacking convention barred them in August, joined by UK Comic Con operator Monopoly Events, British pub chain Wetherspoons, and U.S. Immigration and Customs Enforcement. Courts in England and Wales have also banned smart glasses. An investigation by The Observer and BBC Radio 4 found that a £60 ($78) Chinese model passed through security checks in the UK parliament and could read documents on politicians’ desks.
Meta recently launched a version of its smart glasses without a camera. Meanwhile, the UK government is weighing additional restrictions on smart glasses sales. Verdier said Dutch and European lawmakers should consider the AP’s findings as they craft AI rules, adding, “It is important that we, as a society, learn to deal with these types of high-risk AI products.”
