AINews wire, 1 h ago

Dutch watchdog advises against buying AI smart glasses after safety tests

Reported by BBC News Technology

The Dutch privacy regulator, the AP, has advised people against buying AI smart glasses. It says the devices can get things dangerously wrong. The watchdog stopped short of asking for a sales ban, so the glasses remain available in the Netherlands.

The AP tested six pairs of glasses on the Dutch market. Prices ranged from under €100 (£84) to more than €500, and the devices came from different parts of the world. The regulator has not said which models it tested and will not name them. It is unknown whether Meta's glasses were among them.

In the tests, the glasses gave advice that could be harmful. One pair gave dating advice to an adult who showed romantic interest in a minor. Others misread instructions printed inside medicine packages, and some made errors when interpreting severe weather warnings.

In a simulated emergency where CPR advice was requested, one pair repeatedly said an ambulance was on its way, although emergency services had not been contacted.

The AP also found serious security flaws in some glasses. Someone nearby could connect to them without permission, view recordings, or operate the camera remotely. The regulator also warned that data recorded by the glasses is often sent outside the EU.

The AP raised privacy concerns too. It said the recording light, which shows that a device is filming, could be covered with tape on some models and still work, even though some manufacturers claim it would not. The watchdog's investigation focused mainly on AI and algorithms rather than data protection.

The BBC reports that concerns about smart glasses have grown, including the filming of women without consent and recording in places such as healthcare settings. Meta recently released a version without a camera, and some commentators linked this to the privacy backlash. Meta has said its glasses can help visually impaired people, and the AP agrees they can be valuable for those users. It argues that accuracy matters even more for them. Meta has also said that outsourced workers may sometimes review content to help improve its AI, and that it blurs faces in images first.

What is still unknown: which brands and models were tested, how the manufacturers respond to the findings, and whether any regulator will act beyond this advice. The BBC approached Meta, TCL and HTC for comment. The reports available to us do not include their replies.

Enjoying this? Get the AI brief

Every Thursday: AI that matters for your home.

✓ Free · seven editions · one click to leave

Topics
Story continues on BBC News TechnologyAdd us as a preferred source on GoogleSee our stories first in Google’s Top Stories.
Get alerts on big newsOnly the big news, at most once a day. No email, no account.

Original reporting by the sources above. Summary by smart-home.space.