Meta's AI Glasses Were Sending Private Footage to Human Reviewers Without Wearers Knowing
What happened
Meta's AI-powered smart glasses captured footage and audio from wearers' daily lives and routed that material to human reviewers at a Kenyan contractor called Sama. Workers at Sama reported seeing nudity, sex acts, bathroom use, credit card numbers, and other highly private content in the footage they processed. Some of the people recorded had no idea the glasses were capturing anything at the time, and filters meant to obscure faces sometimes failed, leaving identities visible in the material contractors reviewed.
The arrangement was part of a standard AI development pipeline: human reviewers check model outputs to improve accuracy and flag edge cases. What made this pipeline different was the nature of what smart glasses collect in ordinary life. A device worn through a person's day, at home, in bathrooms, during intimate moments, gathers a fundamentally different category of material than footage a user knowingly submits through an app. Agreeing to wear the glasses is not the same as consenting to have private moments reviewed by a contractor workforce overseas, and the two were not treated as distinct decisions.
When reports about the content and working conditions became public, they prompted lawsuits and regulatory inquiries. Meta paused the human review work and then ended its contract with Sama entirely. The termination affected 1,108 workers in Kenya who had been doing the review work, removing their jobs alongside the pipeline that had created the exposure.
The incident is not simply a story of a corporate arrangement that turned out badly. The pipeline had a structural gap: the people whose footage was reviewed were not informed participants and had no mechanism to find out their material had been seen. The contractors doing the review had no apparent route to refuse particularly sensitive content or to signal that what they were receiving went beyond reasonable review conditions. Both ends of the pipeline operated without meaningful disclosure to the people most affected by it.
That absence is the core accountability failure the incident exposes. No wearer knew whether their interactions had been routed to a human reviewer, when, or who saw the material. No contractor had a formal record of what they had been shown or under what conditions. A provable record of what a system did with captured data, who reviewed it, and what each person in that chain was exposed to, would have made the scale and nature of the problem visible before lawsuits and contract cancellations forced it into the open.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.