Meta ends Kenya AI training contract after workers report exposure to private video content
Workers in Kenya contracted by Meta through outsourcing firm Sama reported in February 2026 that they had been required to review intimate video footage captured by Meta's AI smart glasses, including people undressing and engaging in sexual activity. Meta terminated Sama's contract two months later citing unmet standards, resulting in 1,108 redundancies; a Kenyan workers' group alleged this was retaliation for speaking out, while regulators in the UK and Kenya opened investigations.
What happened
In February 2026, workers in Kenya employed by outsourcing firm Sama reported to Swedish newspapers that they had been required to view videos recorded by Meta's AI-powered smart glasses, including footage of people using toilets and engaging in sexual activity. One worker stated: 'We see everything - from living rooms to naked bodies.'
Meta acknowledged that subcontracted workers reviewed content shared with its AI systems to improve user experience. Less than two months after these reports, Meta ended its contract with Sama, citing that Sama did not meet its standards. Sama denied this, stating it had consistently met all operational, security, and quality requirements and was never informed of any failures.
Reported impact
The contract termination resulted in 1,108 workers at Sama being made redundant. A Kenyan workers' organisation alleged that Meta's decision was retaliation for staff speaking out about their exposure to private content.
Meta smart glasses users had their intimate moments, including undressing in bedrooms and engaging in sexual activity, recorded and subsequently reviewed by workers without apparent knowledge or meaningful consent regarding the specific nature of content they would see.
Index analysis
The UK Information Commissioners Office wrote to Meta describing the reports as 'concerning'. The Office of the Data Protection Commissioner in Kenya announced it was commencing an investigation into privacy concerns raised by the glasses.
The incident highlights tensions between AI training requirements and privacy protections: while Meta stated it had obtained user consent for AI training, the scope and nature of private content reviewers would see appeared not to have been transparent to either users or workers.
Governance controls
Meta stated that humans review AI content to improve product performance and that it receives clear user consent for this practice. However, the revelations prompted regulatory scrutiny from both UK and Kenyan data protection authorities.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.