Meta admitted that it will use all the photos taken by users on Ray-Ban smart glasses to train AI

Meta✴ admitted that any image captured by Ray-Ban smart glasses that users send to an artificial intelligence assistant can be used to train new AI systems.

Image source: ray-ban.com

«In regions where multimodal AI is available (currently the US and Canada), images and videos transmitted by Meta✴ AI may, in accordance with our privacy policy, be used to improve it,” company spokesperson Emil Vazquez told TechCrunch . The company previously stated that photos and videos taken on Ray-Ban Meta✴ glasses are not used by the company to train systems until the user himself sends them to the AI ​​for analysis – at which point the materials fall under a different set of policies. In other words, the company is using its first AI-powered consumer device to create a large data set that can be used to develop even more powerful AI models. The only way to opt out is to simply not access Meta✴AI multimodal systems.

Owners of Ray-Ban Meta✴ smart glasses may not realize that they themselves provide the company with large volumes of images – this could be the interiors of their homes, photos of loved ones and personal documents – to train new AI models. Meta✴ representatives claim that this information is in the user interface of the device, but in reality the company’s management either did not know or did not want to disclose this information. It previously became known that Llama’s AI models are trained on public materials from American users on Instagram✴ and Facebook✴, but now the company has expanded the definition of “public data” to everything that users of smart glasses send to AI for analysis.

The day before, the company began deploying new AI functions for Ray-Ban Meta✴ glasses – communication with the device is becoming more natural, and glasses owners will increasingly send their data to the AI, which the company will use to train new systems. At the Meta✴ Connect 2024 event, she spoke in detail about these new features, but remained silent on what she would do with user data. Meta✴ AI’s terms of service state, “You agree that Meta✴ will analyze these images, including facial features, using AI,” and the company recently paid the state of Texas $1.4 billion to settle a lawsuit related to the system. face recognition. Notably, some image processing-related features of Meta✴ AI do not work in Texas.

Meta✴ also stores transcripts of all voice conversations between users wearing smart glasses by default, also for AI training. But you can refuse to record the voice itself: when you first log into the device management application, the user chooses whether voice recordings can be used for these purposes. By the way, American students have already modified the software of Ray-Ban Meta✴ glasses – now they reveal the name, address and phone number of anyone the user is looking at.