Eye Level

Meta, the company that holds Facebook, came out with the first mainstream "smart glasses" with the launch of their Wayfarer model of Ray-Ban Stories glasses released back in 2021. For the concerning low price of $224, you can get your hands on the Meta AI Glasses, renamed after plastering "AI" on everything gained traction and boosted profits. These glasses contain many enticing features such as Bluetooth support, prescription lenses, a touch sensitive frame, and...

Cameras

All models of the Meta AI Glasses contain two cameras on either side of the frame, both front-facing. These cameras are advertised to be able to take photos on-the-fly and also record video. The cameras do separate things; they aren't combined to make 3D video. One of the cameras does photos, while the other one does video. When recording video, a light appears visible to the people in the line of sight. The cameras are both 12 megapixels and can natively livestream to Facebook and Instagram.

A microphone, too

The Meta AI Glasses also have microphones so that you can use their handy-dandy voice assistant to do certain actions for you only using your voice. You can instruct your glasses to take a picture without even using your hands. You can also start and stop video recording with just your voice. According to Wikipedia, "on April 23, 2024, Meta announced an update to Meta AI on the smart glasses to enable multimodal input via computer vision." This means that they allow the glasses to interact with a Large Language Model (LLM), presumably Meta's Llama LLMs, to identify your surroundings. In order to do that, they need to send pictures from the cameras mounted to your face to their datacenters to process the image and give you a response.

That can't be a privacy problem, right?

You guessed it. You're sending pictures of your surroundings to be processed by Meta directly, a huge data collector. Not only that, but you're potentially sending pictures of other people without their permission. On any given day, multiple people could be sending photos containing you to Meta's servers without your knowledge. It makes it even worse when, as stated by WIRED, Meta is using these glasses to scan for people in the view of them. The article states that Meta discreetly added code to allow for facial recognition in the glasses. They call this feature "NameTag" and it can be used to identify people when they are in the line of view. "NameTag" utilizes 3 different LLMs to detect faces, crop the person out, and translate their information into biometric data. As of yesterday, WIRED reports that Meta has removed the code from user devices and claim that they are not building a database of people's biometric data collected from Meta AI Glasses. It's pretty telling, though, that they already have the code prepared for when they do end up going through with it.