An investigation by Swedish newspapers revealed that contractors in Kenya reviewing footage from Meta's Ray-Ban AI glasses have been exposed to extremely private content, including people undressing, using the toilet, and engaging in sexual acts
Workers report that automatic blurring tools often fail, allowing them to see faces and sensitive information like bank cards, while strict non-disclosure agreements prevent them from speaking out
The UK's Information Commissioner's Office has written to Meta seeking answers.
Meta AI Glasses Showed Sensitive Bank Details, Naked People, Porn To Workers
According to the investigation, some footage appears to have been captured accidentally, such as when wearers place the glasses on bedside tables without realising recording continues.
Summary
An investigation by Swedish news outlets has triggered privacy concerns after revealing that contract workers in Kenya are reviewing highly intimate footage captured by Meta's AI-powered smart glasses—content that includes people undressing, having sex, and sharing sensitive bank information .
The probe by Svenska Dagbladet (SvD) and Göteborgs-Posten (GP) found that thousands of data annotators in Nairobi, employed by outsourcing firm Sama, are tasked with labelling images and videos from Meta's Ray-Ban glasses to train the company's artificial intelligence systems .
Workers described seeing deeply personal moments captured by unsuspecting users. "We see everything, from living rooms to naked bodies," one employee told the Swedish newspapers . "In some videos, you can see someone going to the toilet, or getting undressed. I don't think they know, because if they knew they wouldn't be recording" .
According to the investigation, some footage appears to have been captured accidentally—such as when wearers place the glasses on bedside tables without realising recording continues. Other clips show people watching pornography while wearing the device, or inadvertently filming bank cards and private conversations.
While Meta claims it filters data to protect privacy, including blurring faces, workers told investigators the anonymisation tools frequently fail, particularly in poor lighting conditions. "The algorithms sometimes miss. Especially in difficult lighting conditions, certain faces and bodies become visible," a former Meta employee confirmed to the papers.
The workers operate under strict confidentiality agreements in offices monitored by security cameras, with personal phones banned . Several expressed distress at the content they must review but said economic necessity keeps them in the role. "You are not supposed to question it. If you start asking questions, you are gone," one annotator said .
Meta's terms of service do disclose that "in some cases, Meta will review your interactions with AIs, including the content of your conversations with or messages to AIs, and this review may be automated or manual (human)"
Published At: