Will We Learn to Love the Surveillance Glasses?
— by Isabelle Ulfsdotter
Tags: AI, Privacy, Wearables
Ambient AI wearables may become ordinary without becoming trusted. What would have to be true for constant sensing to become socially permissible?
Disclaimer: For the record, I do have a bias against these devices and I am dismayed by the possibility that they may become commonplace. Nevertheless, I want to attempt to objectively assess the probability of that becoming a reality in spite of my antipathy.
Apple has announced that its new Apple Watch can listen to conversations and summarize them afterward, which seems as good a reason as any to ask whether “ambient surveillance wearables” are about to become part of ordinary life. This includes camera-and-microphone glasses such as Meta’s Ray-Bans and Snap’s Specs; watches capable of capturing, transcribing or reconstructing surrounding conversation such as the new Apple Watch; AI pendants and wearable note-takers; future earbuds or assistants that maintain an ambient memory; and whatever other forms contextual AI hardware assumes (including OpenAI and Jony Ive’s purported “hockey puck” as described by this unconfirmed leak).
We already live among CCTV, traffic cameras, video doorbells and cameras inside shops, offices and restaurants. I am not trying to determine here whether the prevalence of those systems is necessary, effective or proportionate to the security benefits claimed for them. Nor do I assume that they always operate as intended: innocent people can be misidentified, data can be repurposed, and systems installed for exceptional circumstances can become instruments of general surveillance.
Nevertheless, conventional security surveillance is organized around a different premise. It records an environment so that an exceptional event (a theft, collision, trespass or violent act) can be identified or reconstructed. In principle, it is not interested in me as I go about my ordinary life. I become its subject if something happens (or if the system mistakenly concludes that something has happened, or simply misidentifies me).
Ambient surveillance wearables invert that relationship. Their purpose is to make ordinary experience useful to the wearer. When I speak to someone wearing such a device, I am no longer merely an anonymous figure passing through the background of a security recording. I am a protagonist, part of the information the device exists to capture: my face, my voice, my words and the context of our interaction may be interpreted, transcribed, summarized, remembered or transmitted by a private company for someone else’s benefit.
I acknowledge that the boundary is not perfect. Security systems can identify and track ordinary people, while wearables can document genuine wrongdoing. But the organizing distinction remains: security surveillance claims to preserve the ordinary in case something exceptional occurs; ambient wearables make the ordinary itself the object of computation.
I want to conduct a thought experiment by asking:
What would have to be true for ambient surveillance wearables to become common and socially permissible enough that people must routinely expect ordinary interactions with other human beings to be sensed, interpreted or recorded? In other words, “ubiquitous”.
Let me add a further qualification: I do not mean targeted use cases such as someone cycling by, filming a performance or documenting their own activity, where I may appear anonymously as part of the visual or auditory scenery.
1. The utility must outweigh the inconvenience
The devices must be sufficiently inconspicuous and practical that their utility to the wearer > the inconvenience of wearing them.
The practicality part may already have been met. Meta’s glasses look like ordinary Ray-Bans and an Apple Watch is already an ordinary object. Whether the devices are useful enough to wear routinely, rather than for specific activities, is less clear.
Assessment: practicality is probably sufficient; utility remains uncertain.
2. Restrictions must not interrupt their use too much
Workplaces seem like an obvious problem. Companies do not want their intellectual property, customer information or private meetings captured by employees’ personal devices. Nor can I imagine HR departments being delighted by every stupid office remark acquiring its own audiovisual record. Schools have the additional problems of cheating and children recording other children. Hospitals and courts already restrict recording devices for obvious reasons.
Some restrictions are already appearing in Europe. The Dudley Group NHS Foundation Trust asks that recording-capable wearables not be worn in its buildings, and smart glasses have now been banned from courts in England and Wales. Within the EU, the response so far appears to be more regulatory than prohibitory: Irish and Italian data-protection authorities challenged Meta over whether its tiny indicator light gave bystanders meaningful notice that they were being recorded. This is still a patchwork, not a general European ban.
This is where continuity matters. If I cannot wear prescription Meta glasses at work, at my child’s school, in a hospital or during conversations where someone asks me to remove them, will I carry a second pair and put the Meta glasses back on afterward? Restrictions would not make the devices useless. They could, however, reduce them to something people wear for particular activities, which slightly undermines their value as ambient devices.
Assessment: a real obstacle, although probably not a fatal one.
3. Businesses open to the public must tolerate them
Bars, restaurants and cafés feel public, but they are privately owned places open to the public. Their owners can generally impose rules about photography and recording, and some restaurants, pubs and theatres are already restricting Meta glasses.
Imagine two otherwise similar cafés. One allows recording glasses; the other asks patrons to remove them. If enough customers prefer the second because they can speak freely or allow their children to eat a croissant without wondering who is processing their faces, privacy becomes a competitive advantage. The reverse is also possible if nobody cares. Either way, the norm may be shaped by what customers reward.
Assessment: restrictions are emerging, but there is not yet an established norm.
4. People must trust the wearer, the company and the safeguards
For these devices to be accepted, I apparently have to trust three parties at once. The wearer must behave decently. The manufacturer must not misuse the data. The technical safeguards must work.
Meta’s glasses have a forward-facing capture light, and a recent update reportedly disables the camera if the light is covered or damaged. Good. But I cannot inspect the software on a stranger’s face, know whether a device has been modified or tell what is being processed without a conventional recording being saved. Nor do I want a technical discussion about Meta’s tamper protection before ordering coffee.
A 2026 US survey of 1,096 adults found that 72 percent worried about being secretly recorded, 60 percent distrusted manufacturers to handle the data responsibly, and 85 percent supported a visible recording indicator.
Assessment: this condition has plainly not been met.
5. Or trust must become unnecessary
People do not have to approve of a technology for it to become ubiquitous. They may simply get used to it. Acceptance means I think the device is legitimate. Habituation means I stop noticing it. Resignation means I still dislike it but no longer think objecting is worth the effort. To the company selling the device, these may look reassuringly similar.
Flock cameras offer one possible precedent. People object, controversies arise, and the cameras continue to appear. Ambient wearables could arrive through the same gradual process, especially when new capabilities appear as software updates to watches and other objects people already own. There may be no single moment when society decides to accept them.
Assessment: resignation seems considerably more plausible than universal trust.
6. The law must leave enough room for ordinary use
Based on some (quick and superficial) research, it seems that Europe offers more protection than the United States, but not as comprehensive as I would have hoped (what are all those cumbersome regulations for one might wonder). Recording an identifiable person’s image or voice can amount to processing personal data under the GDPR. Outside a genuinely personal or household activity, the person or company doing the processing needs a legal basis and may also owe the subject information about what is happening. The Irish Data Protection Commission specifically describes body-worn cameras as turning the wearer into a mobile surveillance system.
Belgian law is particularly odd. A person who secretly intercepts a private conversation in which they are not participating can commit an offence, but a participant may generally record their own conversation without telling the other participants. What they subsequently do with it can still violate privacy or data-protection law, but the act of recording is not automatically illegal simply because I did not consent.
This leaves a question for watches and other listening devices that continuously process surrounding speech without necessarily saving a conventional recording.
Assessment: European law creates friction, but it does not amount to a clear prohibition on ambient listening or recording.
7. Wearing them must not make you into a creepy weirdo
Google Glass showed that technical utility does not prevent a product from becoming socially repellent. Meta has made its glasses look normal, but their inconspicuousness is precisely what makes them unsettling.
If wearing them in a café comes to mean, “I may be recording you and I have decided that this is your problem,” stigma could suppress everyday use without any law being passed. Dates, parties and family gatherings can become recording-free simply because people refuse to engage until the glasses come off.
Institutional restrictions could reinforce this by teaching people that asking for a recording-free environment is reasonable. The opposite could happen if the devices normalize faster than the objections. The person asking for the glasses to come off may eventually be treated as the difficult one.
Assessment: the social meaning of wearing them is still up for grabs.
8. There cannot be too many spectacular scandals
The possibilities include covert recording, stalking, harassment, biometric identification, leaked conversations, compromised accounts and privacy promises containing an asterisk the size of the aforementioned hockey puck.
Scandal alone does not kill technology. Smartphones and social networks have survived plenty. But a sufficiently vivid incident involving children, patients, colleagues or intimate conversations (or someone being "cancelled" for uncouth statements shared in private) could stigmatize wearers, accelerate institutional restrictions or attract regulators.
Assessment: too early to know, but the category has an unusually large surface area for disaster.
Where does that leave the question?
Some necessary conditions are already partly in place. The hardware is inconspicuous, the functions are improving and surveillance is familiar. The devices can arrive through objects people already wear.
The obstacles are also real. Trust is weak, restrictions are appearing and the social meaning of wearing the devices has not settled. Their success may depend less on people embracing them than on whether objecting becomes too inconvenient.
I would like to conclude that society will reject the entire proposition with the contempt it deserves. I cannot. It may instead normalize it by degrees. The important question is not whether Meta sells a great many glasses. It is whether enough of them enter ordinary human encounters that privacy becomes something I must repeatedly request rather than something I can reasonably assume.
Sources and factual notes
Ambient surveillance: I borrowed the concept from David Lyon’s work on surveillance becoming embedded in everyday life through data trails and background tracking. I am using it somewhat differently and more actively here to mean devices worn on the body which can sense, interpret, record or reconstruct the people and interactions around their owner. See this overview of Lyon’s surveillance studies.