
Hands-free AI And Its Perks
Meta’s Ray-Ban AI Training Work Raises Privacy Concerns
Meta’s Ray-Ban smart glasses send user recordings to human annotators in Kenya to train AI models, prompting scrutiny from EU regulators over privacy and data transfer safeguards.
Meta’s Ray-Ban smart glasses capture content from users and send it for human review to help train AI. Contractors in Nairobi, Kenya have described seeing private moments and sensitive content in the line of duty.
Privacy laws in Europe, especially in countries like Portugal, Sweden, and Ireland, are strict, so regulators are asking Meta to clarify if recordings from Europeans are being sent to Kenya, and whether safeguards meet GDPR (General Data Protection Regulation) requirements.
How the Training Pipeline Works
When users turn on their glasses, they capture images, short videos, and audio clips, usually the sort of thing you'd see walking down the street. That data is sent to annotators overseas, including Nairobi, where workers label and categorize it. Sometimes automatic face blurring or anonymization doesn’t catch everything, so people can still be identifiable in the footage.
Meta says users control what gets shared and promotes the glasses as privacy-conscious, but the terms of service make it clear that some interactions with AI features can be reviewed manually or automatically. Not everyone wearing the glasses, or passing by someone who is, may realize their data is being handled like this.
What's Happening?
Kenyan annotators report labeling sensitive content, including nudity, bank cards, and personal situations that were recorded unexpectedly. The work is usually low-paid and bound by airtight non-disclosure agreements, which is another problem entirely, as gig work conditions in global AI supply chains are questionable.
EU regulators are paying close attention because GDPR requires clear consent before personal data leaves the bloc. Kenya doesn’t have an adequacy agreement with the EU, so Meta has to provide reassurances that European users’ recordings are handled properly.
Hands-free AI has its perks, but someone still has to watch the footage. Right now, that someone might be halfway across the world (~3,400 km in our case). The way AI currently works, convenience often comes with a hidden workforce. If you’re wearing the glasses, your clips could be making a detour before the AI ever sees them.
Tags
Join the Discussion
Enjoyed this? Ask questions, share your take (hot, lukewarm, or undecided), or follow the thread with people in real time. The community’s open, join us.
Published March 4, 2026
published
Latest in Wired World

Meta’s Ray-Ban AI Training Work Raises Privacy Concerns
Mar 4, 2026

Software Engineer-Founded Party Wins 11 Seats in Japan
Feb 23, 2026

India AI Impact Summit Sees Major Investments And Partnerships
Feb 19, 2026

Russia Removes Major Western Platforms From Its Internet Directory
Feb 12, 2026

Africa–Middle East Tech Shows Up Across Web Summit Qatar
Feb 5, 2026
Right Now in Tech

Netflix Drops Out of Warner Bros. Race, Paramount Left Standing
Feb 27, 2026

Court Tosses Musk’s Claim That OpenAI Stole xAI Trade Secrets
Feb 26, 2026

Meta’s Age Verification Push Reignites Online Anonymity Debate
Feb 23, 2026

Substack Adds Polymarket Tools. Journalists Have Questions.
Feb 20, 2026

Netflix Ends Support for PlayStation 3 Streaming App
Feb 18, 2026