Kenyan workers review intimate Meta smart glasses footage
Subcontractors at Meta partner Sama in Nairobi are manually reviewing intimate video clips from Meta Ray-Ban smart glasses, according to a joint investigation by Svenska Dagbladet and Göteborgs-Posten. Workers reported seeing bank cards, nudity, and bedroom scenes recorded without wearers realizing. While retailers claimed data stays local, network tests confirmed the glasses constantly transmit data to Meta's servers for AI processing.

If you leave your smart glasses on a bedside table or wear them into the bathroom, you probably assume the camera is off. An explosive joint investigation by Swedish newspapers Svenska Dagbladet and Göteborgs-Posten reveals that assumption is dead wrong.
Kenyan workers at Sama, an outsourcing subcontractor for Meta located over 9,300 miles from Silicon Valley, are routinely reviewing deeply intimate, real-world footage captured by Meta Ray-Ban glasses. The footage includes users visiting the toilet, partners undressing in bedrooms, banking details, individuals watching porn, and even people wearing the glasses during sex.
The human engine room of multimodal AI
To make Meta's AI understand the real world, someone has to label it. In Nairobi, long rows of data annotators work ten-hour shifts drawing boxes around objects and verifying transcriptions of Meta AI voice chats.
Several workers, speaking on condition of anonymity due to strict confidentiality agreements, described a job that feels increasingly uncomfortable. One employee recalled footage of a man setting his glasses on a nightstand before his wife entered the room to change clothes. Others spoke of people stepping out of the bathroom naked, seemingly entirely unaware the device was recording.
The content is so sensitive that Sama's offices are blanketed in security cameras, with workers strictly banned from bringing personal phones or recording devices into the building.
What retail salespeople promise vs. what the network logs
The investigation also exposed a glaring disconnect between retail sales pitches and technical reality. The Meta Ray-Ban line surged in popularity in 2025, reaching an estimated seven million units sold in partnership with EssilorLuxottica, up from two million in 2023 and 2024 combined.
Reporters visited ten eyewear retailers across Stockholm and Gothenburg, including major chains like Synsam and Synoptik. Sales staff consistently reassured undercover journalists about data privacy. One Synsam employee insisted that nothing is shared with Meta and that users have full control, while another salesperson claimed everything stays stored locally within the mobile app.
The technical test told a completely different story. Reporters bought a pair, went through setup, and explicitly opted out of sharing additional data to improve products. Working alongside a system developer, they analyzed the app's network traffic.
The result: smart functions refuse to work offline entirely. To interpret what the camera sees or answer questions, the app continuously transmits data to Meta server hubs in Luleå, Sweden, and Denmark. When asked about the discrepancy, retail chains responded that ultimate responsibility for adhering to local laws and Meta's terms falls squarely on the wearer.
Why it matters
This investigation cuts straight to the central paradox of first-person AI wearables. Devices like Meta Ray-Ban are sold as friction-free digital assistants, but right now, lightweight onboard silicon cannot handle complex multimodal reasoning. Everything has to ping the cloud, and running an effective AI pipeline still relies heavily on low-cost human annotators reviewing raw inputs behind closed doors.
For consumers, the takeaway is stark: toggling off product improvement settings does not keep your video feed on your device if you interact with the assistant. Furthermore, bystanders and family members never agreed to Meta's terms of service, yet they are the ones being flagged in bedrooms and bathrooms.
We think this revelation will sharply accelerate regulatory pushback against always-on facial cameras across Europe and beyond. Until on-device edge processing can handle vision models without cloud handoffs, the smartest privacy setting for any pair of camera glasses remains physically packing them into their case.
Sources (1)
- svd.se(opens in a new tab)Originalsvd.se