Kenyan technology workers tasked with reviewing footage captured by smart glasses developed by Meta Platforms and Ray-Ban have reportedly been exposed to deeply personal moments recorded by users around the world, according to a recent international investigation.
The revelations have reignited debate about privacy, artificial intelligence training, and the often-hidden human workforce powering modern AI systems.
The investigation found that Kenyan workers employed by outsourcing firm Sama review video clips recorded through the camera built into the Ray-Ban Meta Smart Glasses.
Their role is part of a process known as data annotation, where human reviewers label objects, scenes, and activities appearing in images and videos so artificial intelligence systems can learn to interpret the world more accurately.
The smart glasses allow users to capture photos, videos, and audio hands-free while also interacting with built-in AI features designed to recognise objects and answer questions about the environment.
To improve the technology, some recordings are reportedly sent to human reviewers who verify or correct the AI’s interpretation.
But workers involved in the process say the footage sometimes contains far more than everyday scenes.
According to accounts from annotators, the clips occasionally include highly sensitive and private moments that users may not have realised were being captured.
Workers described seeing people: using the bathroom, getting dressed or undressed, engaging in intimate activities, having private household conversations, displaying personal information such as bank cards or documents.
One reviewer reportedly said the job sometimes involves viewing scenes from people’s homes and bedrooms, moments that are typically never intended for outside audiences.
The glasses, developed through a partnership between Meta and Ray-Ban’s parent company EssilorLuxottica, allow users to record short video clips and interact with Meta’s AI assistant.
Some of this content is used to train and refine the AI system. In certain cases, clips are shared with human reviewers who verify whether the AI correctly identified what appears in the footage.
Unlike smartphones, smart glasses can record from the wearer’s perspective without drawing attention, raising concerns about whether bystanders, or even the user themselves fully understand what may be captured.
In many cases, individuals being recorded may have no idea they are appearing in footage used to train AI systems.
The latest revelations come as governments and regulators worldwide push for greater transparency around artificial intelligence and data collection.
