Do AI glasses see too much?

Good morning. Let’s make sure today’s headlines don’t surprise you tomorrow. 😌

What’s up? An investigation found that Meta’s AI-powered smart glasses have sometimes recorded very private moments — like people getting dressed or using the bathroom — and some of that footage has been reviewed by data workers overseas.

Want the details? Meta’s Ray-Ban smart glasses include a built-in camera, microphones, and an AI assistant that activates when users say “Hey Meta.” The system processes things the glasses capture, such as speech, images, and sometimes video, to help improve the AI assistant. According to reports, some of this data is sent to contractors who review the recordings and label them so the AI can learn. Workers at a third-party company in Kenya said they have occasionally seen highly personal footage, including people getting dressed, going to the bathroom, or revealing sensitive information like bank cards. Meta says this review process is part of improving the product and is described in its privacy policy.

Why should you care? This finding highlights growing privacy concerns as wearable AI devices become more common. Glasses that can record what users see raise questions about how much personal information is captured and who ends up seeing it. Even if companies say data is filtered and protected, the idea that human reviewers might view private moments is undoubtedly worrisome.

What’s up? Researchers at Anthropic introduced a new way to measure how much AI could affect different jobs — and early data suggests AI has not yet caused widespread unemployment.

Want the details? The researchers created a metric called “observed exposure,” which looks at both what AI can theoretically do and what people are actually using AI for at work. They combined job-task data from the O*NET database with real-world AI usage data to see which occupations are most affected. Jobs like computer programming, customer service, and data entry currently show the highest exposure. However, the study also found that AI is still being used for only a fraction of the tasks it could theoretically perform. When researchers compared these results with employment data, they found no clear increase in unemployment in highly exposed jobs since late 2022.

Why is this significant? Job loss because of AI is very real, and this is not meant to discredit those who have been impacted. Instead, the findings suggest that AI’s impact on jobs may be slower and more complex than many people expect. Instead of sudden job losses, the early effects may appear in subtler ways, such as slower hiring — especially for younger workers entering AI-exposed fields. This new measurement system could help researchers track how AI changes the labor market over time and identify which jobs may face disruption before major shifts become obvious.

Hi — I’m Hunter, a PhD candidate whose work has appeared in major academic journals and popular tech outlets. I founded FryAI to make staying ahead of AI clear, accessible, and fun.