An interesting and in-depth report from the Stockholm daily Svenska Dagbladet, recommended reading aided by the fact that it was written in English and not Swedish, once again puts the theme of “privacy” at the center with reference to the operation of Meta’s smart glasses, an American multinational known mainly for having Facebook, Instagram, and WhatsApp in its portfolio but increasingly active in the wearable device segment, having also acquired a minority stake (about 3%, with valuations up to 5% under study) in the well-known European eyewear manufacturer EssilorLuxottica for about 3.5 billion dollars.
WHAT DOESN’T ADD UP ABOUT META SMART GLASSES
As is known, the Ray-Ban Meta and subsequent versions are sold because they promise and allow you to always have at your disposal a set of information and functions related to the things framed by the lenses. What seems less known is their practical operation, since what is captured by the cameras is not limited to being processed “on site,” by a sort of electronic brain installed in the glasses but may be sent to the company’s servers for video processing first and then for a response.
The glasses constantly record short clips every time the “real-time AI” is activated. During these sessions, the device’s camera and microphone remain online so that the Artificial Intelligence can analyze the scenes and respond to the user’s questions. The data is then uploaded to Meta’s systems, where it becomes part of a vast sample used to refine future versions of the digital assistant.
DO SMART GLASSES IMAGES END UP IN KENYA?
According to what has been reconstructed, at least some of the data would end up in Nairobi, Kenya, where human reviewers from the unspecified African company Sama, by virtue of a complex mechanism of contracts and subcontracts probably set up to minimize costs, review the footage and train the AI to avoid mistakes.
If what has been documented is also confirmed regarding the nature of the videos and audio viewed and listened to by human technicians, it is first of all clear that the average consumer is unaware of the complex journey around the globe undertaken by the footage recorded with smart glasses. Even less aware can be those who, not wearing Meta’s smart glasses, still find themselves within their range and see their image, including biometric data, transported to the other side of the world.
Also because most users probably think the wearable device’s operation is linked to the onboard AI’s ability to recognize what is framed by the lenses and do not think that these images are instead viewed by human “suggesters” who have the thankless task of assisting the Artificial Intelligence to avoid glaring mistakes.
ARE META GLASSES PRIVACY-PROOF?
Meta, for its part, believes it operates correctly: everything would be within the terms of service and the AI privacy policy. These documents would inform the consumer that content recorded by Ray-Ban Meta and company is sent to servers and can be assigned to human reviewers for model improvement, but since they are often written in legalese not understood by most of the population, it remains to be asked whether those who wear them really know how data and images are processed. Furthermore, the Swedish newspaper argues, the documentation does not offer many details on the treatment and management of sensitive content.
Indeed, as highlighted by TechSpot, Kenyan collaborators stated they have examined footage of people using bathrooms, dressing, and in some cases even having sex, all recorded from the perspective of the glasses. And since Meta’s smart glasses not only record video but also capture sounds, the audio of some clips would include discussions about protests, criminal activities, or deeply personal aspects of people’s lives, all elements that are unclear who listens to and views after being transmitted to Meta’s servers.
