Meta's AI glasses send intimate videos to human reviewers in Kenya

Meta has spent years assuring users that its smart glasses are safe. Now it sits at the center of yet another privacy scandal, and this time the scale of the problem looks genuinely alarming. An investigation by two Swedish newspapers, Svenska Dagbladet and Göteborgs-Posten, found that video captured by Ray-Ban Meta cameras is sent to real people for review: company contractors working in Nairobi, Kenya. The material they have seen includes recordings from bathrooms, sex scenes, and other deeply intimate moments from the lives of unsuspecting users.

To see why the investigation struck such a nerve, it helps to recall the context. Meta launched its smart glasses with Ray-Ban back in 2023 and has consistently stressed that the device is "designed with privacy in mind." An LED on the frame was supposed to signal to bystanders that recording was underway, and the company repeatedly said it handles user data with the utmost care. Those promises became a core marketing pitch: without them, selling a camera built into an everyday accessory would have been much harder. It now turns out that reality looks radically different from the ads.

Sending user data to contractors in low-wage countries for moderation is nothing new in tech. Similar scandals have already hit Amazon over Alexa, Apple over Siri, and Google over Google Assistant. In each case, the companies admitted that humans listened to snippets of recordings, supposedly to "improve service quality." But the Meta case is fundamentally different: this isn't about audio clips of voice commands, but about video that captures a person's daily life from a first-person view. The camera on the glasses records everything the wearer sees, and, as it turns out, that "everything" includes the most private moments imaginable.

The technical explanation lies in how the glasses' AI features work. When a user activates the Meta AI voice assistant, the device can capture video and send it to the company's servers for processing. Some of those recordings end up with moderators — people who label data to train machine learning models. The problem is that users apparently don't grasp the extent of what is being transmitted, or to whom. As is often the case, the wording in the user agreement is vague enough to give the company legal cover without giving users any real understanding of what's happening.

The backlash came quickly. In response to the Swedish report, at least one class action lawsuit has already been filed in the US, accusing Meta of violating false advertising and data protection laws. The plaintiffs point specifically to the company's marketing claims that the glasses are "designed for privacy" — a claim that, after the investigation, looks like outright consumer deception. If the court sides with the plaintiffs, the financial consequences for Meta could be significant, although for a company worth hundreds of billions of dollars, fines rarely act as a serious deterrent.

Far more important is the reputational damage and its long-term impact on the entire wearable AI device industry. Meta isn't the only company betting on smart glasses and first-person cameras. Snap, Google, and dozens of startups are working on similar products. Every scandal like this erodes consumer trust not in a single brand but in an entire class of devices. People are already wary of cameras in everyday objects — just remember the fate of Google Glass, which failed largely because of public rejection. If the industry can't offer genuinely transparent and verifiable data protection mechanisms, wearable AI devices risk remaining a niche product for enthusiasts.

The Kenyan side of the story deserves special attention. Nairobi has long been one of the largest hubs for outsourced content moderation. Contractors for Meta, OpenAI, TikTok, and other companies work there for pay that is modest by Western standards, doing work that global AI systems depend on. Journalists have previously documented psychological trauma among moderators forced to view violent and explicit content every day. Now there's another layer: people in Kenya are watching intimate recordings of users in Europe and the US who don't even suspect these viewers exist.

The scandal confronts the industry with a fundamental question that has no good answer yet. Modern AI systems need human data labeling — it's an unavoidable part of training. But where is the line when it comes to first-person video shot during the most private moments of someone's life? Meta will have to find an answer fast, because the next generation of its glasses is rumored to get even more advanced AI features and even more powerful cameras.

*Meta is designated an extremist organization and banned in Russia.