A Nairobi-based data firm said it has reviewed sensitive footage captured by Meta’s Ray-Ban smart glasses after the tech giant tapped the Kenyan company for AI training offshore.
“In some videos, you can see someone going to the toilet, or getting undressed,” an unnamed source told reporters last week. “I don’t think they know, because if they knew, they wouldn’t be recording.”
John Davisson, Deputy Director of Enforcement at the Electronic Privacy Information Center, said the technology raises broader concerns about how wearable devices collect and use personal data, particularly in a public setting.
“The wearer of the glasses cannot consent on behalf of all of the people they are encountering as they go through the world using these glasses,” Davisson told Decrypt. “Whether in public places or in private places, at places like locker rooms, restrooms, or other intimate spaces.”
Davisson also said training AI systems on that footage increases risk because the data can include identifiable faces, voices, and other personal information.
“You are compounding the privacy and data protection concerns, because you're taking people's personal information and using it to build your own model,” he said.
Davisson said reports of the glasses recording people in intimate situations do not surprise him and suggested companies may attribute such recordings to false activations or other technical explanations.
“But the fact remains that they are capturing sensitive information that no reasonable consumer would want their smart glasses to capture,” he said.
“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),” per Meta’s terms.
Meta claims it cannot read or access shared messages when a user disseminates private information with friends, family, and AIs using processing technology.
Still, the company may utilize user content and related information through automated systems, human review, or through third-party vendors to improve its services and conduct research.
It may also conduct research on a user’s content to ensure compliance with its own policies and applicable laws, while removing content that violates its rules, the company claims.
Svenska Dagbladet and Göteborgs-Posten’s investigation identified the Nairobi-based data subcontractor for Meta as Sama, which employs workers in Nairobi to train AI systems by manually annotating data, including videos, images, and speech, for Meta’s AI services.
“Every image must be described, labelled and quality assured,” the report said. “All to make the next generation of smart glasses a little more intelligent, a little more human.”
Workers told the Swedish newspapers they reviewed footage that included people using the bathroom, changing clothes, credit card numbers, and explicit sexual activity.
“There are also sex scenes filmed with the smart glasses. Someone is wearing them, having sex. That is why this is so extremely sensitive,” a contractor told reporters. “There are cameras everywhere in our office, and you are not allowed to bring your own phones or any device that can record.”
Contractors also told the newspapers they felt unable to question the assignments for fear of losing their jobs.
“When you see these videos, it feels that way. But since it is a job, you have to do it,” another said. “You understand that it is someone’s private life you are looking at, but at the same time, you are just expected to carry out the work. You are not supposed to question it. If you start asking questions, you are gone.”
Meta did not respond to a request for comment by Decrypt.
