Meta Explained to Me the Privacy Policies of Its AI Smart Glasses. I’m still worried

I dress up Meta’s Ray-Bans turn off and on when I’m walking to take pictures, take calls and listen to music. Technology is interesting, fun and useful.

I also knew that Meta’s privacy policies could be a concern, but now I’m more concerned about it than ever.

My concern grew after a number of friends and colleagues shared a report about Meta’s third-party contractors in Kenya being able to view sensitive information such as images of bank records, nudity and sexual acts that were recorded on Meta’s glasses (resulting in a class-action lawsuit).

What restrictions did Meta put in place to protect people’s privacy? I checked Meta’s terms of service online and the Meta AI app, but that wasn’t helpful.

I wanted answers. So I contacted the Meta comms team for clarification.

But even after getting an official answer from Meta about where the lines are drawn, I’m still confused and unsure. Although most people are justifiably worried about someone secretly recording themselves with smart glassesThere’s another wrinkle: When will these glasses be able to share what you’ve been recording?

Here’s a short answer: Does Meta Mirrors have third party contractors who may be looking at your data? Yes, sometimes — if you use AI services. If you don’t use those AI services, according to Meta, you should be fine. But still, I don’t know where that wall of “AI services” is clearly drawn. And that is one of my biggest concerns.

Meta has had a long history of problems with both privacy and trust, spanning over the past decade and Cambridge Analytica scandal. Those problems did not arise Meta’s VR headsetwhich don’t have many AI resources to collect data, but the company’s smart glasses do. And those services will continue to grow and become more powerful over the next few years. Ray-Ban’s popular Meta sunglasses — more than 7 million pairs sold last year — are the best the whole wave of camera-enabled AI glasses and wearables from many companies, with Google enters the mix later this year.

If you’re interested in Meta glasses, which, as a technological breakthrough, are the best camera and audio-enabled smart glasses at the moment, you need to keep this concern in mind. And as smart glasses evolve into AI-enabled devices, we’re going to face a lot of questions about how comfortable you can feel relying on their services — and what cloud-based AI technology companies should do to make these policies clearer.

Below, I’ll share Meta’s answers at length so you can understand my reasoning​​​​​​​​​​​​​​​​​​​​ and and make your own assessment about the risks.

Using AI resources with Meta Ray-Bans

If you use AI — for example, to analyze something you see or get a translation — then third-party contractors may look at what you record.

Here’s what the company told me: “Ray-Ban Meta glasses help you use AI, hands-free, to answer questions about the world around you. Unless users choose to share media they’ve captured with Meta or others, that media stays on the user’s device.”

But there is this: “When people share content with Meta AI, we sometimes use contractors to review this data for the purpose of improving people’s experiences, as do many other companies. We take steps to filter this data to protect people’s privacy and help prevent identifying information from being reviewed.”

The takeaway from this is that whenever you use Meta’s AI services, Meta may be using third-party contractors to update the data.

Although Meta promises that the information is properly filtered to remove sensitive data or information, that worrying news report said that the Kenyan contractors were describing photos taken on glasses with sensitive images clearly visible.

That worries me more about what happens when people use Meta AI for assistive purposes: that is, as a way to “see” when you can’t see with your own eyes. Can looking at personal documents and reading back be a dangerous thing to do? Since Meta hasn’t exactly introduced any kind of encrypted, secret AI features to its glasses, it’s possible.

Meta has this to say about protecting privacy: “We have strong policies and surveillance rules that intentionally limit what contractors can see.”

But again, I really don’t know what those strict policies or guidelines are for monitoring.

“We take steps to filter this data to protect people’s privacy and help prevent identifying information from being reviewed,” Meta added.

This does not help clarify any details. I’m going to be honest here, which is totally wrong.

I have to assume that anything done with cloud AI services, like using Meta, can be seen to some extent by third party contractors. And you should too.

Taking photos and videos with Meta’s Ray-Bans

Meta glasses don’t always use AI, and neither do I. In fact, I use the Meta glasses to record photos and video, listen to music, and make phone calls. I don’t use AI much, in part because Meta’s AI has very little interaction or control over my other personal data or my iPhone.

With non-AI photo and video recording, things should be safe… I guess.

I asked members of the comms team if the photos or video recordings I made with the glasses, and which were not involved in the AI-based request, could be subject to viewing by third-party contractors. They said: “To be clear, the photos and videos users take with their AI glasses are stored in the camera on their phone. not used by Meta to develop and improve AI. If you just record a video or take a photo using the mirror camera button, that media stays on your phone. Unless you choose to share media you’ve captured with Meta or others, that media stays on your device.”

That sounded promising. But with the Meta glasses settings, the storage gets a little cloudy… literally. In the Glasses privacy settings for the Meta AI app, the Cloud Media toggle says it “allows your photos and videos to be sent to the Meta cloud for processing and temporary storage.”

Will cloud media mean that my personal photos and videos are open to annotation by third-party contractors? According to Meta, no. According to Meta, any commands that use AI to send images or use AutoCapture methods enabled by changing Cloud Media will also be safe.

In the company’s words: “Certain features, such as sharing your glasses using your voice (‘Hey Meta, send a photo’), automatic seamless media import, or Auto Capture, where the camera takes photos or videos automatically when you start the feature (useful in times when you may want to capture content without manually triggering the camera using a button or voice), may need to send you temporary storage or your videos uploaded to the cloud. to cloud media services, photos and videos sent from in frames or automatically imported to your phone are not subject to human annotation.

Meta does not clearly define what exactly “Cloud Media” is, other than a temporary storage place for your photos and videos to be processed with voice commands. And what worries me is how the wall is drawn around “private” media versus “AI-connected” media. It makes me want to turn off Cloud Media, which means that the photos and videos are stored in my phone’s photo library.

What to do with AI glasses now?

I still love the camera and audio features of the smart glasses and I’m intrigued by the upcoming AI features. But I’m also very concerned about the uncertainty about where the line is drawn between what is attributed to a third party, which may be, and what remains confidential. Meta uses those third parties to help train AI, or rate content. It’s a reminder that many AI services are cloud-based and out of our control.

I get even more worried when I think about the Meta search reports add facial recognition and more on its smart glasses.

Meanwhile, more AI glasses are coming, and wearable camera-equipped AI devices, too. Google will be next. And all of these companies need to make it more clear how they use data from these devices, how they protect our privacy concerns, and how we users can manage it — if at all. It is not at all easy to understand how Meta mirrors handle AI data, or where it is sent. I hope this story helps you better understand where the lines might be.

However, I have to admit that I feel very unlikely to use Meta Mirrors for anything personal or data sensitive. Travel glasses? A tool for quick social photos for work, however I broadcast? Testing with AI? I think so.

But if Meta intends to be a tool that helps us deeply with AI wearables, and doesn’t want everyone to call them “distorting glasses,” which people already are, it needs to do better, fast.