Dutch watchdog: don't buy AI glasses. One said CPR help was coming.
After testing six models priced from €100 to €500, the Dutch Data Protection Authority calls the devices 'unsafe' for users and bystanders, and tells policymakers the trust gap is now a safety problem.

The Dutch Data Protection Authority advised consumers against buying AI-powered smart glasses after testing six models and finding safety failures, including one that falsely told a wearer emergency services had been called during a CPR simulation, plus weak recording lights and unauthorized access by nearby people. For hardware makers, enterprises, and policymakers, the finding converts a privacy debate into a product-safety and disclosure liability problem in one of Europe's largest markets.
The Netherlands' privacy regulator has a blunt message for shoppers: do not buy AI-powered smart glasses. The Dutch Data Protection Authority issued that advice after testing six models, priced from under €100 ($110) to over €500 ($550), and finding failures that go well beyond the recording habits that made the category controversial in the first place. In a CPR simulation, testers watched one pair of glasses tell the wearer that emergency services had been called and were scheduled to arrive at the scene. Nobody had made the call. "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," said Monique Verdier, vice chair at the AP.
Safety was not the only casualty. The AP said one pair offered dating advice to an adult wearer who expressed an interest in a minor. It did not name the model, though a spokesperson confirmed that behavior was linked to a €354 pair, and declined to map specific failures to price points, saying the research was meant to flag risk across the category rather than score individual products. Two of the six were "extensively tested in simulated real-life situations where human safety played a key role." Others mishandled safety warnings on children's toys, misread medication leaflets, and misinterpreted warnings for extreme weather events. If you wear glasses to check a pill packet or to catch a heatwave alert, that is a consumer-protection problem wearing a consumer-electronics costume.
The warning is not a first strike. In September, the AP reminded the public that everyone has a right to privacy and encouraged organizations to ban the devices where necessary, flagging risks around data processing under the GDPR and around the training of AI. This week's tests doubled down: the AI functions rely on processing data through servers mostly located outside the European Union, and users were not given sufficient information about how the AI was processing the data they captured. "AI offers many opportunities, but we must be able to trust that technology," Verdier said. "With the camera glasses examined, that trust cannot currently be justified."
Then there is the security layer, which reads like a list of everything a hardware startup hopes nobody audits. Testers found that people near a wearer could connect to the device without permission, view recordings, and in some cases operate the camera. With "virtually all of the camera glasses examined," the recording light meant to signal an active capture was dim and easy to cover, and models could be induced to keep recording even with that indicator hidden, contradicting what manufacturers so often claim. "As a result, bystanders can barely see whether they are being filmed," the AP said. That is the mechanism behind the devices' colloquial nickname, "pervert glasses," a label these findings will not help shake.
None of this erases the legitimate upside. Smart glasses can assist the visually impaired, people with lower literacy levels, and anyone who needs help understanding the world around them. But the watchdog's official position on the technology as it exists today is that it is "unsafe... due to the risks to users and bystanders." Verdier called on politicians in the Netherlands and across Europe to weigh the results when developing AI policy. "It is important that we, as a society, learn to deal with these types of high-risk AI products," she said. "Simply banning everything provides some leeway in the beginning, but we cannot sustain that indefinitely. It is important that we collectively work on awareness, resilience, and ethical conscience."
The backlash is already institutional. DEF CON announced a no pervert glasses policy ahead of its August event, a move broadly welcomed by attendees, including EFF director of cybersecurity Eva Galperin, who said at the time: "Love to see a 'no pervert glasses' policy at DEF CON." Monopoly Events, which manages UK Comic Cons, pub chain Wetherspoons, and even Immigration and Customs Enforcement have imposed similar restrictions. The physical security argument got harder to dismiss after a joint investigation by The Observer and BBC Radio 4 found a £60 ($78) pair of Chinese-made glasses slipped past the airport-style scanners used in the UK parliament and could be used to read upside-down documents on MPs' desks, recording data such as phone numbers and addresses for later access.
For the executives building, selling, insuring, or deploying this hardware, the AP's findings reframe the risk. Public recording via smartphone is already normalized, and on paper glasses are not wildly different. The difference is legibility: a phone in hand broadcasts intent through body language, while smart glasses look like analog eyewear and, per the AP, can hide their own recording indicator. That gap is where liability lives. A national regulator in one of Europe's largest markets has now formally advised consumers against the product category, turning a privacy argument into a safety and disclosure question for manufacturers, a compliance question for enterprises weighing deployment, and a political question for anyone drafting AI rules. The fix is not obviously technical or cheap. Brighter and tamper-resistant indicators, better consent flows, clearer data-processing disclosures, and auditable on-device behavior are the baseline, and the AP's verdict is that today's models are not there. Verdier's framing leaves the door ajar: banning is easy early, unsustainable later. Companies that move first on trust may find that door open for them. The ones that do not should expect more regulators to walk through it.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology
Anthropic mathematician claims Claude helped crack the Riemann hypothesis
If the proof holds, AI has toppled math's most famous open problem; if it does not, it is a case study in announcing breakthroughs before anyone verifies them.
AI's $4 Trillion Wipeout Erases Nasdaq's Gains for the Year
A brutal week took roughly $4 trillion off the AI sector and left the Nasdaq flat for the year, and the damage ran the full length of the supply chain.
Anthropic mathematician says AI proved the Riemann hypothesis. Verification will take time.
A single unverified claim from inside Anthropic now sits against more than 160 years of open mathematics, and the gap between asserting a proof and surviving peer review is the entire story.


