Sweating through his shirt, a man sits opposite two journalists in a small Nairobi hotel room, fearing that his employer will discover his presence. In essence, his work entails viewing footage taken with Meta’s Ray-Ban smart glasses, which occasionally shows people handling credit cards, undressing, or using the restroom. He did not volunteer to be a voyeur. However, Meta created him for that purpose.
Earlier this year, a joint investigation by the Swedish publications Svenska Dagbladet and Göteborgs-Posten revealed the story, which went viral in the already contentious wearable AI debate. Employees in Kenya who were hired to examine and annotate video for Meta’s AI training pipeline reported being exposed to extremely personal footage, recordings that the subjects most likely were unaware were being shared. A few of the clips featured nudity. Financial data was recorded by others. This was not requested by the employees. The subjects of the film didn’t either.
Meta has consistently used a legalistic defense. According to the company’s terms of service, users consent to the possibility of human review. Technically, that is accurate. Additionally, it’s the kind of disclosure that is so obscured by legalese that almost no one reads it, much less comprehends its ramifications. Consenting to a vague clause is not the same as realizing that a stranger on a different continent might be watching you leave the shower.
This is especially disturbing because of the scale. Zuckerberg has referred to the seven million pairs of Meta Ray-Bans that have been sold thus far as “some of the fastest-growing consumer electronics in history.” The glasses are nearly identical to standard Ray-Bans. The tiny light on the frame that serves as the recording indicator is so dim that it is invisible during the day. When they discovered what their glasses were recording and where that footage ended up, even some wearers were taken aback. People who were unaware that they had been recorded and those who were unaware that their recordings had been sent for review have already filed two lawsuits.
In the meantime, the social repercussions are already apparent. Men wearing Meta’s glasses are surreptitiously filming women at beaches and in public areas, frequently for content that is uploaded without permission. When a woman asked a creator to take down a video of her, she was informed that it was “a paid service.” It’s difficult to ignore how well this technology, repackaged with a sleek frame and a Silicon Valley sheen, fits into preexisting harassment patterns.
There are more privacy issues than just creeps on the boardwalk. The ACLU, EPIC, and Fight for the Future are just a few of the civil liberties organizations that have called on Meta to scrap plans for a facial recognition feature known as “Name Tag.” The fear is simple: allowing millions of people to covertly identify strangers in public would be a gift to immigration enforcement officers, abusive ex-partners, and stalkers. Meta might consider this to be a game-changing feature. Advocacy groups consider it to be more akin to a weapon.
Nevertheless, the market continues to expand. According to reports, Apple is creating its own smart glasses. Following the spectacular failure of Google Glass ten years ago, Google is getting ready for a second attempt. An updated version of Snap’s specifications is being made available. Within a few years, researchers predict that 100 million people will wear AI glasses of some kind. This raises concerns about whether courts, hospitals, theaters, and restrooms can actually enforce recording bans against a technology that is meant to be invisible.
There is a perception that the industry has determined that the privacy debate is merely a speed bump that should be handled more by PR and terms-of-service updates than by actual product design. “The onus is ultimately on individual people to not actively exploit” the technology, a representative for Meta told the BBC. This is an amazing statement about a product whose whole design philosophy makes detection almost impossible and exploitation simple. The Nairobi employee realized something that the company’s messaging carefully avoids: privacy isn’t a setting you can change once the camera is always on. It’s something that has already vanished.
