Behind the world’s largest social media and artificial intelligence platforms is a global workforce responsible for reviewing some of the internet’s most graphic material, including depictions of beheadings, child sexual abuse, torture, and suicide.
Accounts published by the Data Workers’ Inquiry, a research initiative backed by academic and technology institutes in Europe and the U.S., describe moderators in Kenya reviewing traumatic content under strict productivity quotas while struggling with psychological distress they said often went untreated.
What content moderators do
Social media companies rely on content moderators to review graphic posts, videos, images, and comments that may violate platform rules or break the law.
While automated (AI) systems remove some material, companies still rely heavily on human reviewers to determine whether content contains hate speech, graphic violence, child sexual abuse material, or other prohibited content.
Moderators review queues of flagged posts and decide whether material should remain online or be removed. Their decisions are also used to help train artificial intelligence systems designed to detect similar content automatically in the future.
A workforce built on outsourcing
Content moderation is largely carried out by contractors in lower-income countries rather than by the technology companies that operate the platforms.
One account said Meta, which has 2.9 billion users and generates approximately $2.5 billion in revenue per day, contracted the outsourcing firm Sama to handle moderation work in Nairobi, Kenya. Sama employed thousands of workers there on content moderation and other AI-related data work for Meta and OpenAI, the report said.
According to accounts published by the Data Workers’ Inquiry, workers earned about $1.50 per hour, or roughly 40,000 Kenyan shillings per month after taxes, according to the account. Workers were required to act on each piece of content, known as a “ticket” within 50 seconds.
“I was no longer a person,” a moderator identified by the pseudonym Yaro said in a report by Kauna Ibrahim Malgwi, who worked as a Hausa-language moderator for Meta. “Just a machine doing tasks every 50 seconds.”
‘I thought I was in hell’
By routine, workers encountered material depicting beheadings, torture, child sexual abuse and mass atrocity. Workers described severe psychological effects they said frequently went unaddressed.
“I thought I was in hell,” a worker identified as Khan Rafiq said in Kauna Ibrahim Malgwi’s report. “I saw people getting beheaded every day.”
Workers interviewed in the accounts described using prescription medications, alcohol, and other substances to cope with the psychological effects of moderation work. One worker described a suicide attempt that allegedly occurred on the production floor.
“After work I sought refuge in a bottle of heavy alcohol,” a worker identified by the pseudonym Lethabo Lubanzi said in one report. “This job had left me with so many addictions.”
The reports said Sama provided on-site counselors, but several workers questioned their qualifications and said they believed counselors shared information with management.
“Not certified psychologists or real professionals,” a worker identified as Jandyose Mukasa said in one account.
Malgwi, who said she was later diagnosed with post-traumatic stress disorder, anxiety, and depression after working as a moderator, proposed a trauma-informed mental health program for groups of data workers. Her proposal cited the March 2025 death of a Nigerian colleague as evidence of what she described as an urgent need for greater psychological support.
Meta has stated that it takes its responsibility to content reviewers seriously and conducts regular independent audits of partners and encourages reviewers to raise concerns. It has emphasized that contractors like Sama are responsible for day-to-day employment conditions. Sama has defended its practices and called many allegations inaccurate or misleading while noting content moderation was a small part of its business.
Lawsuits challenge Meta’s outsourcing system
The Nairobi content moderation operation has become the subject of multiple lawsuits challenging how outsourced content moderators were recruited, paid, and treated.
One case, filed in Kenya by a former moderator, alleges that workers recruited from other African countries were not fully informed about the nature of the job before relocating to Nairobi. The lawsuit alleges job advertisements failed to disclose that employees would be reviewing graphic Facebook content, including depictions of violence and child sexual abuse material and the alleged concealment violated Kenyan labor and anti-trafficking laws.
The case also alleges that the worker was fired after attempting to organize workers and form a union at Sama.
The lawsuits remain ongoing. According to Foxglove, settlement negotiations between Meta, Sama, and the former workers broke down in late 2023. A ruling expected in February 2026 was postponed without a new date being set, and petitioners said they were considering asking Kenya’s chief justice to intervene, according to Capital FM Kenya.
Foxglove has said the cases could have broader implications for the global content moderation outsourcing industry.
Looking ahead
Last month, Meta ended its contract with Sama, resulting in Sama laying off over 1,100 workers in Nairobi, after determining that the partner “did not meet our standards,” citing a specific prior incident uncovered in a February Swedish investigation where Sama workers in Nairobi had been reviewing raw, unblurred video and image footage from users’ Ray-Ban Meta smart glasses to train AI models.
The footage reportedly included highly private and intimate content such as people using the toilet, undressing, sexual activity, bank card details, and everyday home moments — material that many users appeared unaware was being recorded or sent for human review. Meta emphasized that such photos and videos are private to users and that any human review for AI improvement occurs only with clear consent via its privacy policy, adding that it requires all vendors to maintain strict security, privacy safeguards, and compliance standards.
The company said the incident contributed to its decision to transition the work to other partners as it shifts toward more advanced AI systems.