Meta has recently launched its AI-powered Ray-Ban smart glasses, which feature a discreet camera on the front for taking photos. Notably, these photos are not only taken upon request but can also be captured automatically when certain AI functions are triggered, such as when you say "look." This implies that the Ray-Ban Meta glasses will collect a large number of photos, both those taken by the user and those captured unintentionally.
Recently, a TechCrunch reporter inquired with Meta about whether they plan to use photos taken by Ray-Ban users to train AI models, similar to how they use images from public social media accounts. Meta did not provide a clear response to this query.
Anuj Kumar, Meta's Senior Director of AI Wearables, stated in a video interview: "We do not discuss this publicly." Meanwhile, Meta spokesperson Mimi Huggins also noted: "This is not something we typically share externally." When asked by TechCrunch if Meta was training on these images, Huggins responded: "We neither confirm nor deny it."
This is particularly concerning, as Ray-Ban's new AI features are set to be activated for extensive passive photography. Last week, TechCrunch reported on Meta's plan to introduce a real-time video feature, which, once activated by specific keywords, would stream a series of images (essentially real-time video) into a multimodal AI model, allowing for low-latency, natural responses to questions about your surroundings.
This means that users may inadvertently capture a large number of images in their daily lives and upload them to the cloud-based AI model. For example, when a user asks the glasses to scan the clothes in their wardrobe to help with outfit selection, the glasses will actually take dozens of photos of the room and its contents.
However, Meta is reluctant to disclose how these photos are processed after being uploaded.
Wearing Ray-Ban Meta glasses essentially means you are wearing a camera on your face. As we learned from the experience of Google Glass, the acceptance of facial cameras by others is not high. Therefore, companies typically launching such products should clearly state: "Hey! All photos and videos taken by your facial camera are completely private and for your personal use only."
Yet, Meta has not done this. Meta previously announced that they would use AI training on all public posts from U.S. users on Instagram and Facebook. The company considers all this information as "publicly available data," and we may have to accept this reality. Their definition of "publicly available" is very broad, blurring the boundaries of user privacy.
Although we cannot confirm whether Meta is using images captured by the Ray-Ban Meta camera to train AI models, the company has also not explicitly denied this.
Other AI model providers have clearer regulations on this matter. For example, Anthropic states that they never use customer input or output data for training, and OpenAI also states that they never use customer input or output data from the API for training.
Key Points:
📸 Meta's Ray-Ban smart glasses come with a hidden camera capable of automatically capturing photos of your surroundings.
🤔 Meta has not clearly stated whether they will use photos taken by users to train AI models, raising concerns about privacy.
🔒 Other AI companies have stricter restrictions on the use of user data, making Meta's policies appear vague.