BTA2023 An investigation by Swedish newspapers Svenska Dagbladet and Göteborgs-Posten has revealed that Kenyan contract workers reviewing data for Meta Platforms are exposed to highly personal content captured by users of the company’s AI-powered smart glasses.
The report sheds light on the hidden human labour behind wearable AI systems and raises renewed concerns about privacy, cross-border data transfers, and the psychological toll of digital moderation work in Nairobi.
Behind the AI: Human Review in Kenya
The workers are employed by Sama, a Nairobi-based outsourcing company that provides data annotation and content moderation services to major global tech firms.
According to interviews cited in the investigation, Kenyan reviewers label and analyse photos and videos captured through Meta’s Ray-Ban smart glasses, which were developed in partnership with EssilorLuxottica.
The glasses are marketed as AI assistants capable of:
- Translating languages
- Describing surroundings
- Capturing hands-free photos and videos
- Answering contextual questions
However, many AI features require cloud processing, meaning user content may be transmitted to remote servers and reviewed by human annotators to improve system performance.
“I Don’t Think They Know”
Several Kenyan workers reportedly told investigators they regularly encounter sensitive or intimate footage, including:
- Recordings inside bedrooms and bathrooms
- Users undressing or engaging in private activity
- Financial details such as visible bank cards
- Personal domestic scenes unintentionally captured
One worker was quoted as saying:
“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.”
Another claimed they reviewed footage where a wearer left the glasses on a bedside table, unintentionally recording a spouse entering the room and undressing.
These accounts raise questions about user awareness, consent, and how transparently AI wearables communicate data usage to consumers.
Data Protection and GDPR Questions
For European users, Meta’s Irish subsidiary oversees compliance with the EU’s data protection framework. However, Kenya does not hold an EU adequacy decision recognising its data protection standards as equivalent to the General Data Protection Regulation (GDPR).
The investigation therefore highlights broader concerns about:
- Cross-border data transfers
- Transparency in AI data training
- The scope of human review
- Safeguards around sensitive material
Meta’s privacy policies acknowledge that user content may be subject to human review to improve product performance and safety.
The Human Cost of AI Training
Sama has previously faced scrutiny over labour conditions in content moderation contracts, including earlier partnerships involving other AI systems. Workers in Nairobi’s annotation sector often report:
- Low wages
- High productivity quotas
- Exposure to disturbing or explicit material
- Strict confidentiality agreements
While AI products are marketed as automated and seamless, the investigation underscores the extent to which human reviewers remain central to machine learning development.
A Blurred Line Between Automation and Oversight
Meta has previously defended its data practices, stating that it invests in privacy safeguards and minimises the amount of data used for AI training.
Still, the findings suggest that wearable AI technology may collect more intimate moments than many users realise — and that those moments may travel thousands of kilometres for human analysis.
As AI-powered devices become more embedded in daily life, the episode raises fundamental questions about privacy, ethical AI development, and the often invisible labour powering global tech innovation.
