Meta Ray-Ban smart glasses have a small LED light on the frame that blinks when the camera is recording. According to Meta, this light is the privacy feature, a design solution for the fact that you are now wearing a camera on your face in public.
I find myself thinking about that light. Not because it is inadequate, though it is. I think about it because it invites interpretation, like all urban objects do. The stoplight, the security camera, the construction fence: Interpretation is how we move through cities, though we are seldom aware of it. The LED on the Ray-Ban frame is one such signal. The problem is that nobody really understands what it means.
Recently, Swedish journalists broke news that should have brought the news cycle to a screeching halt. Meta pays subcontractor Sama, based in Nairobi, Kenya, to manually label videos taken by Ray-Ban smart glasses. A user says “Hey Meta” to the artificial intelligence on the glasses in asking it to examine what it’s seeing. That then sends the video to Meta’s servers, then it goes on to workers who label what they see on the video: objects, faces, locations, etc. These workers, the journalists found, see people undressing, using the bathroom and recording their own bank cards by accident. One annotator described the work as: “We see everything.”
Seven million pairs of these glasses were sold in 2025. Each pair generates training data. Each “Hey Meta” command routes footage through human eyes on two continents, where they work for low wages under nondisclosure agreements, surveillance cameras and a strict no-questions policy. Workers who raised concerns about what they were seeing were fired.
This is the same Sama that Time magazine exposed in 2023 for paying Kenyan workers less than $2 an hour to label graphic violence and abuse content for OpenAI. Workers described the experience as psychologically devastating, while contracts stated that OpenAI would pay an hourly rate of $12.50 per worker per hour.
Meta technically discloses this arrangement buried in its privacy policy. A sentence notes that the company reserves the right to conduct “manual (human) review” of AI interactions. That sentence is doing enormous legal work. It is the LED light of the legal world. Technically visible but designed not to be seen.
Urban planners have a concept called legibility, developed by Kevin Lynch in the 1960s. The idea is that cities should be designed so residents can orient themselves, understand their environment and feel a sense of agency within it. A legible city is one where the street grid makes sense, where landmarks anchor neighborhoods, and where you know where you are and roughly how things work. An illegible city produces anxiety, alienation and disorientation.
The smart-glasses problem is a legibility problem.
When you walk down Michigan Avenue today, you cannot tell who around you is recording. The LED light blinks only when active, and even if you spotted every active device, you would have no way of knowing that your image, caught accidentally and without context, might be reviewed by a worker in Nairobi who has no choice but to keep watching no matter what appears on screen.
This is not a hypothetical privacy risk. It is an active data pipeline running through one of the most documented failures of AI labor ethics on record, operating at scale in every city where 7 million pairs of glasses are being worn.
Cities are negotiated spaces. The compact is implicit but real: When you walk outside, you accept a degree of visibility. You might be seen, photographed or remembered. But that compact has always assumed some rough parity. The person seeing you is also visible to you, subject to the same social norms and standing on the same sidewalk. The smart-glasses pipeline breaks that entirely. The person reviewing your footage is invisible to you, working under rules you never agreed to, in a city you have never visited, for a company that paid $2 an hour for the privilege.
Chicago is not a naive city when it comes to surveillance. We have more cameras per capita than almost any American city. The city has debated, legislated and litigated cameras in public space for decades. But those debates assumed a fixed infrastructure. Cameras mounted on poles, operated by identifiable institutions, subject to Freedom of Information Act requests and City Council oversight. Wearable AI cameras are none of those things. They are mobile, privately operated, legally obscured and outsourced to labor markets with no connection to the cities being recorded.
Meta is already working on the next generation of these glasses. Facial recognition, the ability to identify strangers on the street in real time, is coming. The company that cannot reliably blur faces in training data wants to start naming them on purpose.
The LED light will still be there. It will still blink. Most people still will not know what it means.
Cities are legible when their systems are honest about what they are and how they work. A security camera mounted on a pole has an owner, an operator and a legal framework. You can dispute it, submit a FOIA request about it and protest it. A tiny light on a fashionable frame, signaling nothing interpretable to the people walking past, connected to a labor pipeline that runs from the Magnificent Mile to a Nairobi office where workers are fired for asking questions: That is not a privacy feature. It is a design decision made to look like one.
That is a city that no longer knows how to read itself.
Yunus Emre Tozal is a civil engineer in Chicago and a master’s of art student at Catholic Theological Union.
Submit a letter, of no more than 400 words, to the editor here More Top Picks Best Folding Treadmills