On October 5, Norway’s Ministry of Digitalisation and Public Governance issued a press release in Norwegian that, within a day, had spread through news agencies worldwide: the government intends to submit a bill to the Storting for a temporary ban on “KI-briller”—smart glasses with cameras and AI—in certain public places.
Digitalisation Minister Torgeir Mikkelsen summed up the motive in one sentence: he was “uncomfortable with the fact that a powerful new technology is being introduced in places where people risk being photographed, filmed, or recorded without knowing it.” From that cautious ministerial phrasing grew the season’s most radical proposition. Previous restrictions on smart glasses were written by those who controlled a venue: a pub chain and a cinema operator, each for themselves. Oslo plans to write the ban into statute for parks, beaches, and schoolyards—places that have no house rules and never could.
The list of places is curious in itself. Parks, beaches, museums, shopping centres, public events. Separately—everything built for children: schools, kindergartens, playgrounds, youth clubs, with universities still an open question. And entirely separately—places where privacy matters more than usual: medical facilities, doctors’ offices, swimming pools, gyms, anything with changing rooms and showers.
There is no total ban. Private use remains legal, as does wearing them in places where the owner of the glasses does not risk recording an unwilling person. The bill is promised “as soon as possible,” while the government is simultaneously assembling an expert group to work on permanent regulation—described in the release as “wearable technology.”
An important caveat, easy to miss: the ban is planned with exemptions for vulnerable groups, so that “socially beneficial use” remains possible. That is not a figure of speech. Meta is giving 15,000 such glasses to blind adults in Ireland, and that programme works precisely because of the camera. A Norwegian parliamentary document explicitly states that the ban need not conflict with assistance for people with disabilities if an exemption is written in.
The rush is easy to explain: the market is growing faster than the law. EssilorLuxottica, the maker of Ray-Ban, reported selling more than 7 million AI glasses in 2025—against 2 million across 2023 and 2024 combined. In September 2025, a model with a display in the lens and gesture controls launched for $799, and Bloomberg wrote in January that Meta and EssilorLuxottica were discussing at least doubling production to 20 million by year’s end.
The first objection that comes to mind: everyone has a camera phone in their pocket, so how are glasses worse? The difference is in how data protection law works. It regulates data processing, not the fact of surveillance itself. A camera owner in a park can formally record all they like—the violation begins with publication, not filming.
Smart glasses break that model because you cannot see them at work. In guidance issued on September 22, the Dutch Data Protection Authority stated bluntly that a recording light is insufficient to inform people about the processing of their personal data—which means the mechanism Meta spent all summer reinforcing resolves nothing legally. It also contains a line that kills the convenient argument of “I’m not saving anything”: even when a photo or video is not stored anywhere, personal data may still be processed by AI functions.
France’s CNIL took the same thought one step further in June, toward its social consequence. The shift from fixed, regulated, recognisable video surveillance to mobile, nearly invisible, ubiquitous surveillance creates in people a habit of constantly wondering whether they are being recorded—and that is a form of self-censorship, striking at freedom of expression, assembly, and protest. The Norwegian parliamentary document says the same thing from the other side: people will start avoiding demonstrations or fearing to express their own identity because they are afraid of being identified.
The most unresolved part of the proposal is the definition of the device. The press release honestly lists three options and chooses none of them: (1) glasses with a camera and audio recording—a purely hardware definition, encompassing even a $99.99 pair from a smartwatch brand, regardless of whether some assistant is looking through the lens; (2) glasses with a camera and AI functions—a software definition, where the same frame falls under the ban or falls outside it depending on what a firmware update has enabled; (3) “other wearable technology with a camera, audio recording, and/or AI”—and here, with that “and/or,” the wording spreads to pendants, pins, and watches, including glasses without cameras but with a microphone and an assistant.
Three options mean three different laws, and Oslo has published a question instead of an answer. California stumbled at precisely the same point: five days earlier, Governor Gavin Newsom vetoed SB 1130, a law punishing the recording of people without permission in changing rooms and doctors’ offices, precisely because the definition of a wearable recording device might have swept in smartwatches.
The second unresolved question is what counts as “use.” Taking a photo in a changing room is an easy case. The hard one looks like this: the wearer of the glasses stands in a park and asks the assistant what kind of tree that is; the app sends a single frame to a model. If places are regulated by category, this situation will sooner or later turn the question “take the glasses off or not” into a requirement for the platform to know where it is. Drones went through this process: their no-fly zones began as rules a pilot had to check before takeoff, and a couple of product cycles later became map data in the aircraft’s own firmware. No one has proposed that for glasses.
The first foray into this territory came thirteen years ago. In March 2013, the London group Stop the Cyborgs released posters that any establishment could use to announce: “Google Glass is banned on these premises.” The posters were downloaded thousands of times, and Seattle’s 5 Point cafe banned the glasses before they even went on sale. The activists’ wording is revealing: “We are not calling for a blanket ban. We want people to set social and physical boundaries around the use of technologies themselves, rather than fatalistically accepting the direction they are heading in.”
A little over a year later, consumer Google Glass was quietly shut down. And in 2017, the device returned under the name Glass Enterprise Edition—now in plants and factories, where consent is obvious because everyone on the shop floor is on the same side. The lesson of that cycle is simple and dispiriting: society tried to work it out through etiquette, etiquette failed, and the product left public space for places where the question of consent does not arise. Norway’s bill is an attempt to travel the same road, but straight through statute, because thirteen years of argument have produced not a single rule that can be posted at the entrance to a park.
While lawmakers argue over definitions, physics offers its own answer. The open-source Ban-Rays project is building a hidden-recording detector from two independent methods.
The first exploits a property of CMOS sensors: any camera sensor reflects infrared light back toward its source—the very “cat’s eye” that gives away any optical system in the dark. An Arduino Uno setup sends pulses from infrared LEDs at 940 and 850 nanometres, while a photodiode reads the reflection. A camera sensor responds with sharp, rapid spikes; ordinary reflective surfaces produce long, broad waves. Early tests on Ray-Ban Meta produce unstable results at a distance of around ten centimetres.
The second method goes after the radio signal: Ray-Ban Meta glasses broadcast a specific manufacturer identifier, 0x01AB, and service UUID 0xFD5F over Bluetooth Low Energy. It can only be caught at the moment of startup or pairing; to see the signal during normal operation, modules for sniffing connection packets are needed—so far, that remains in the project’s plans.
In the Hacker News discussion, the idea was immediately picked apart. One witty counterexample: a detector with its own camera is “a good guy with a gun against a bad guy with a gun.” The project’s author replied that he would feel stupid if his solution used its own camera, and that the no-optics “probing” method works unreliably. Another point was made separately: glasses looking for other people’s cameras may have their own IR emitter visible in the recording made by that very observer.
Perhaps the longest-running dispute is over the tiny light in the corner of the frame. Meta insists that the tamper-resistant LED flashes white during recording, that the camera shuts off if someone tries to cover or damage it, and that “this does not exist on smartphones or other cameras.”
Hamburg’s data protection commissioner took the glasses apart and wrote in a report dated September 10, 2026, that the LED is “not easily visible in many circumstances, particularly outdoors and in direct sunlight.” There is another detail there that explains much about the second generation: when asking about an object or person nearby, the first-generation LED “glows very faintly,” while on the second generation it does not light up at all.
Then the market takes over. An Engadget journalist bought a $16.99 sticker kit—a transparent film with slits plus a black sticker on top. The slits let exactly enough light through to the sensor that the protection does not trigger, while the diode itself remains covered. Outdoors, the light is completely invisible; the warning does not trigger even once. The option for Oakley sports glasses is a solid plastic cover, $15 for ten, fitted like an official accessory. Meta responded by promising to look for ways to better detect tampering, remove listings for such “accessories,” and block offenders’ accounts, up to and including bans. Instagram head Adam Mosseri had previously said the company would ban those who post videos harassing people through modified glasses.
Norway is not the first to move—and not the fastest. In the Netherlands, everything was settled in ten days without a single law. On September 22, the regulator issued guidance for venue owners; on September 29, the consumer association Consumentenbond gave Meta two weeks to alter the design and default settings, promising a lawsuit otherwise. On October 1, online retailer Wehkamp removed the glasses from sale; on the 2nd, optical chain Hans Anders did the same, including more than 140 stores in Belgium.
The pressure behind that decision is revealing: almost none. In written answers to parliament dated September 24, the government reported that the regulator had received six signals and one complaint, called a partial ban on wearing them in public places a far-reaching measure, and preferred to wait for the conclusions of the European Data Protection Board. A week later, two retailers removed the product on their own. Distribution can shut down for reasons absent from every statute.
The market data, meanwhile, conflict in the details, and that is worth remembering. The Guardian writes that “at least 9 million pairs of Ray-Bans” have been sold since the beginning of 2023; EssilorLuxottica’s reporting speaks of 7 million AI glasses in 2025 alone, including Oakley. Different ends of the same calculation, and both figures circulate in the press as if they were interchangeable.
There is another discrepancy—over the LED. Meta claims it cannot be covered: the camera will shut off. The Norwegian parliamentary document says the light is easy to cover, disable, or simply fail to notice, while detecting a recording is left to the individual person. Formally, both are right: you cannot break it, but you can cover it without the sensor noticing—and kits for doing that are sold for the price of lunch.
Until there is a text before the Storting, the whole undertaking has only one hard fact: the proposal does not say exactly what is prohibited. The three candidates for its definition—hardware, software, and “and/or”—lead to three different laws and three different lists of those affected: from $99 glasses to microphone pendants. This bill will have to be read starting with the definitions clause, and if it chooses the software option, the lifespan of the ban will be measured in firmware updates.
Everything else is an argument about whether the right not to be recorded can be protected by technical means at all. So far, it is going badly: physics catches hidden optics at a distance of ten centimetres, an indicator light is extinguished by a $2 sticker, and a law that lists categories of places will sooner or later become map data in someone’s firmware. The dry answer to the question posed by the hook is this: the right to go unnoticed has already been lost at the hardware level, and the only question is how expensive society is prepared to make that loss in legal terms.