Meta has admitted that images captured by Ray-Ban smart glasses, when sent to its AI assistant, can be used to train new AI systems. This has raised concerns among users about privacy and data usage. According to Meta spokesperson Emil Vazquez, in regions where multimodal AI is available (currently the US and Canada), images and videos sent to Meta AI may be used to improve the technology in accordance with the company’s privacy policy.
Previously, Meta stated that photos and videos taken with Ray-Ban glasses were not used to train AI systems unless the user explicitly chose to send them for analysis. Once sent, the materials fall under different policies, which allow Meta to use them for AI development. Essentially, the company is leveraging its consumer AI device to build a vast dataset that can further enhance its AI models. The only way to avoid this is to refrain from using Meta AI’s multimodal systems.
User Data and AI Model Development
Ray-Ban Meta smart glasses owners may not fully realize that they are providing Meta with a large volume of images, potentially including personal spaces, loved ones, and sensitive documents, which are used to train AI models. While Meta claims this information is available in the device’s user interface, it appears that the company’s management either overlooked the clarity of this disclosure or chose not to highlight it prominently.
Meta’s use of user-generated content for AI training has expanded over time. Previously, Llama AI models were trained using public materials from platforms like Instagram and Facebook. Now, the definition of “public data” seems to have extended to include any content smart glasses users send to the AI for analysis. At the recent Meta Connect 2024 event, the company detailed new AI features for Ray-Ban glasses, making communication with the device more seamless. However, Meta remained silent on how it plans to handle the increasing volume of user data. We’ll keep you updated on further developments.
Voice Recording and User Modifications
In addition to images and videos, Meta also stores transcripts of voice conversations between users and smart glasses by default, using them for AI training as well, notes NIX Solutions. Users can opt out of this feature during the initial setup of the device management app, where they can choose whether or not to allow their voice recordings to be used for AI purposes.
Interestingly, American students have already modified the software of Ray-Ban Meta glasses to reveal sensitive information, such as the name, address, and phone number of individuals the user is looking at. This raises further privacy concerns regarding the potential misuse of data collected by the device.