Dutch regulator advises against buying AI smart glasses after safety tests

The Dutch Data Protection Authority tested six models and found faulty advice in safety scenarios, weak recording indicators and security gaps that let nearby people connect without permission.

· 2 min read

black-framed eyeglasses
Photo: Claudio Schwarz / Unsplash

The Dutch Data Protection Authority (AP) has advised consumers not to buy AI-powered smart glasses, according to The Register. The advice follows tests that found problems well beyond the privacy concerns the devices are already known for.

The regulator examined six models priced from under 100 euros to more than 500 euros. Two of them were tested at length in simulated real-life situations where human safety was at stake. The AP did not name the models, and a spokesperson declined to tie particular behaviors to particular price points, saying the aim was to show the risks of AI glasses in general.

Several results concerned the quality of the AI's answers. One pair offered dating advice to an adult wearer who had expressed interest in a minor. In a CPR simulation, another told the wearer that emergency services had been called and were on their way, although nobody had made a call. Some glasses did not handle safety warnings on children's toys properly, some made errors when reading medication leaflets, and some misread warnings about extreme weather.

The AP also repeated earlier concerns about data handling. Its tests showed that the AI features mostly depend on servers outside the European Union, and that users were not given enough information about how their captured data was processed. On security, people close to a wearer could in some cases connect to the glasses without permission, view recordings and even operate the camera.

Bystander privacy was another focus. With virtually all of the camera glasses examined, the lights meant to show that recording is underway were dim and easy to cover. The AP said some models could still be made to record with the light covered, which contradicts what manufacturers often claim. As a result, it said, bystanders can barely tell whether they are being filmed.

The regulator acknowledged useful applications, such as helping people who are visually impaired or have lower literacy levels. Even so, it currently regards smart glasses as unsafe because of the risks to users and bystanders. AP vice chair Monique Verdier said that with the camera glasses examined, trust in the technology "cannot currently be justified." She urged politicians in the Netherlands and across Europe to take the results into account in AI policy, and said that blanket bans cannot be sustained indefinitely and that society needs to build awareness and resilience.

The warning builds on one the AP issued in September, when it reminded the public of the right to privacy and encouraged organizations to ban the devices where necessary. The Register noted that DEF CON announced a policy against such glasses ahead of its August event, and that other organizers and organizations have introduced similar restrictions. It also pointed to a joint investigation by The Observer and BBC Radio 4, which found that a 60 pound pair of Chinese-made AI glasses passed through the security scanners at the UK parliament and could be used to read documents on MPs' desks.

Related