The Future of Smart Glasses: Privacy Concerns and AI Integration

Meta has not disclosed whether it trains AI using images captured by its Ray-Ban smart glasses. These glasses, equipped with a front-facing hidden camera, can take photos both on command and autonomously when activated by certain keywords like "look." This raises concerns about privacy since the glasses may collect numerous images, both intentionally and unintentionally, without a commitment from the company to keep these images confidential.

When asked if it plans to use images from Ray-Ban smart glasses to train AI models, similar to its practice with publicly shared social media images, Meta did not provide a response. Anuj Kumar, a senior director for AI wearable devices at Meta, stated in a recent TechCrunch interview that such information is typically not shared with the public.

The apprehension surrounding this issue is heightened by a new AI feature for the glasses, which will enable them to capture numerous passive images. TechCrunch recently reported that Meta intends to launch a real-time video feature for the Ray-Ban smart glasses. When activated by specific keywords, the glasses will stream a series of images (effectively live video) to a multimodal AI model, allowing it to respond to questions about the surrounding environment with low latency.

This collection of images raises the question: what happens to these photos afterward? Meta remains silent on this matter.

Wearing Ray-Ban smart glasses means having a camera on your face, which, as seen with Google Glass, can make others uncomfortable. One might expect a company in this space to assure users that "Hey! All your photos and videos from your front-facing cameras will be completely confidential and isolated."

However, that is not the stance Meta has taken. The company has previously announced that it trains its AI models on public posts from Instagram and Facebook users, categorizing this data as "public." This broad interpretation of what constitutes public data for AI training contrasts sharply with the private nature of the world viewed through smart glasses.

While there is no definitive proof that Meta trains its AI models using footage from the Ray-Ban cameras, the lack of clear communication leaves users in doubt. In contrast, other AI providers, like Anthropic and OpenAI, have stricter policies regarding user data, explicitly stating they do not use customer input or output for training their AI models.

We have reached out to Meta for further clarification and will update this story if they respond.