There's a question few people ask themselves when they put on a pair of Ray-Ban Meta glasses: who else is seeing what I see?

The answer, according to recent reports from researchers and journalists who have reviewed Meta's moderation policies, is uncomfortable. The videos captured by these smart glasses —fragments of everyday life, the inside of a home, moments with family, routine errands— are reviewed by human moderators in Nairobi, Kenya, as part of the training and moderation of Meta's AI models. This isn't an algorithm processing images. These are real people watching what the user sees.

That shifts the privacy conversation. We're no longer talking about abstract data on remote servers. We're talking about human eyes that could catch a glimpse of a credit card number at a store, see a couple or children in personal moments, or wander through someone's home from an intimate vantage point. The same model applied to Facebook's content moderators in Africa —workers hired cheaply to handle difficult material— is now being extended to first-person video of daily routines.

Ray-Ban Meta glasses are marketed as a stylish accessory with tech capabilities. They capture photos, record video, stream live, and answer queries with AI assistance. In the advertising, Meta describes them as instruments of freedom: hands-free capture of life without complications. What's left out of the ads is the infrastructure behind it. Any AI model for conversation or computer vision needs data to learn from and human oversight to function. That oversight has a definite geography: centers in Kenya and other African countries.

These moderation centers are no secret in the tech industry. Meta, Google, TikTok and other platforms have shifted moderation tasks to lower-wage economies. In 2023, TIME published an investigation into OpenAI's moderators in Kenya, who handled sensitive content for less than two dollars an hour. The logic is economic: companies create value in hubs like Silicon Valley and distribute the human costs where they're cheapest. It's not a conspiracy. It's a structure that raises ethical questions.

What sets Ray-Ban apart is the kind of content it generates. With a phone, there's a natural pause: you take it out of your pocket, activate the camera, decide to record. With glasses, that pause disappears. The camera is always ready, always in the field of view, always potentially active. Users capture —sometimes without realizing it— scenes they wouldn't record with a phone: private conversations, intimate moments, sensitive documents, everyday payments. The device's design eliminates the friction that makes us hesitate before recording.

Technology models that simplify user interaction almost always benefit the model more than the individual. Recording becomes easy, more gets recorded, more data is produced, the AI model gets refined, the product becomes more appealing. From an engineering standpoint, it's an efficient cycle. From the perspective of a user who has no idea their financial information is passing before eyes on another continent, it's something else entirely.

Meta and similar companies argue on the grounds of consent: users agree to the terms of service. Formally, that's correct. In practice, it lacks depth. The terms for Meta's hardware devices span dozens of pages of dense legal language. Few people read them in full. They're designed to be accepted, not understood. When consent stems from documents that require legal training to parse, it becomes a formality for dodging accountability.

This recalls debates over contracts of adhesion in the nineteenth century, when railroad and insurance companies used standard forms that customers signed without negotiating. It took courts decades to correct those abuses. With digital technology, we're repeating the process, but faster than regulators can respond. The European Union has made progress with the GDPR and the AI Act, though implementation is moving slowly. In Mexico and Latin America, the frameworks are even more fragile.

The core problem goes beyond personal privacy. It's about power imbalances. Meta has access to what you see. You don't know who reviews it, how it's used, or what gets derived from it. That one-way flow of information sustains the business models of the largest tech platforms. We've seen it with ChatGPT and advertising, with TikTok and children's data, with AI chats lacking legal safeguards. The trend doesn't change: human intimacy as a resource.

The geography of this operation says it all. The data is born in countries with high purchasing power, where the devices are affordable. Processing happens in nations with low labor costs. Profits flow back to shareholders and executives in tech hubs. This chain reproduces inequalities documented for centuries. Technology doesn't dissolve them; it dresses them up in modern interfaces.

I'm not sure how to address this comprehensively. Regulations are necessary, but not sufficient on their own. Technical solutions —robust encryption, on-device processing, decentralized AI models— exist, though the major platforms have no incentive to adopt them. Information remains the first line of defense: knowing that the glasses aren't entirely private, that convenience carries a hidden cost, that the 'magic' involves human labor somewhere.

Before buying a device like this, the relevant question isn't its style or its features. It's who else has access to what it captures, under what conditions, and what happens to that information afterward. Questions no advertisement will answer.

Stones don't lie, but historians sometimes do.

Sources:

1. TIME Investigation: "OpenAI Used Kenyan Workers on Less Than $2 Per Hour" (2023) — documents the conditions of outsourced moderation in Kenya

2. Meta Terms of Service for Ray-Ban Smart Glasses — data use and content moderation policies

3. European Data Protection Board — guidelines on wearable devices and personal data processing

4. The Guardian: "Content moderators at TikTok, Facebook and YouTube" — pattern of outsourcing moderation to emerging economies

5. Electronic Frontier Foundation (EFF) — analysis of consent in digital adhesion contracts