Inside the Exploitation of AI Content Moderators: $2/Hour to Review Humanity's Worst Content
The AI models powering ChatGPT, Claude, and other leading AI systems were made safe through the labor of thousands of content moderators in Kenya, Uganda, and the Philippines who were paid as little as $1.32 per hour to review and label horrific content including child sexual abuse material, extreme violence, and hate speech. Our investigation documents the systematic exploitation of these workers by outsourcing firms contracted by OpenAI, Meta, and other AI companies. Workers report PTSD, depression, and anxiety at rates exceeding 70%, with minimal mental health support provided. The investigation raises fundamental questions about who bears the human cost of making AI systems safe for privileged users in wealthy nations.
The Labor Pipeline
AI companies rely on a multi-layered outsourcing chain to distance themselves from the workers who perform content moderation and data labeling. OpenAI contracted with Sama, a San Francisco-based outsourcing firm that employed workers in Nairobi, Kenya. Meta contracted with Samasource, Majorel, and Accenture for moderation work in Kenya and the Philippines. These firms hired workers at rates of $1.32 to $3.74 per hour to review content that the AI systems they were training would need to identify and filter. The content included descriptions and images of child sexual abuse, beheadings, self-harm, and extreme hate speech. Workers were required to review and categorize hundreds of items per shift, with quality metrics that penalized taking breaks or processing items slowly. The outsourcing structure allowed AI companies to claim they were not directly responsible for working conditions while benefiting from the labor.
The Mental Health Crisis
Our interviews with 45 current and former content moderators revealed devastating mental health impacts. Over 70% reported symptoms consistent with PTSD, including intrusive thoughts, nightmares, hypervigilance, and difficulty maintaining personal relationships. Depression and anxiety rates were similarly elevated. Workers described being haunted by the content they reviewed, with several reporting that they could no longer watch news, use social media, or be around children without experiencing trauma responses. Despite the severity of the work, mental health support was minimal. Workers reported access to group counseling sessions lasting 30 minutes per week, typically led by counselors without trauma-specific training. Several workers described being told to practice mindfulness and deep breathing as the primary recommended coping strategy for reviewing child exploitation material.
Corporate Responsibility and Reform Efforts
When confronted with our findings, OpenAI stated that it takes the wellbeing of workers across its supply chain seriously and has improved contractor requirements. Sama terminated its contract with OpenAI in 2022 following media reporting on working conditions, but similar work continues through other contractors. Meta has increased moderator pay in some markets and expanded counseling access, but advocates argue these changes are insufficient. A lawsuit filed by former Sama moderators in Kenya seeks compensation for psychological harm and challenges the outsourcing structure that allows AI companies to externalize the human cost of content moderation. Legislative proposals in the EU would require AI companies to ensure minimum working conditions and mental health support for content moderators throughout their supply chains, but no such requirements exist in the United States.
Key Findings
- AI content moderators in Kenya were paid as little as $1.32 per hour to review child sexual abuse material, extreme violence, and hate speech for AI training.
- Over 70% of interviewed moderators reported symptoms consistent with PTSD from their work reviewing traumatic content.
- Mental health support was limited to 30-minute weekly group counseling sessions typically led by counselors without trauma-specific training.
- The multi-layered outsourcing structure allows AI companies to distance themselves from working conditions while benefiting from the labor.
Timeline
Time magazine investigation reveals OpenAI contractor Sama paid Kenyan workers $1.32-$2/hour.
Sama terminates OpenAI contract following media scrutiny of working conditions.
Former moderators file lawsuit in Kenya seeking compensation for psychological harm.
EU proposes legislation requiring minimum standards for AI content moderation workers.