Summary
Long-form journalism synthesizing the global “ghost work” economy that underpins AI systems — data labelers, content moderators, and annotation workers who do the human labor that AI is marketed as replacing. Documents the geographic concentration of exploitation in Kenya, the Philippines, India, and Venezuela; the corporate subcontracting architecture that enables plausible deniability; and the emerging organizing resistance from workers.
Key Points
- Every major AI system requires enormous quantities of human-labeled data; this is not a temporary phase but a permanent structural dependency
- Workers earn $1–2/hour (median hourly wage on Amazon Mechanical Turk: $2; only 4% earn more than $7.25/hour)
- In Venezuela: 90 cents to $2/hour for data labeling; same work pays $10–25/hour in the US
- Scale AI (valued at $29B) coordinates hundreds of thousands of global workers for Meta, OpenAI (until recently), and the US military; takes profit while avoiding employer responsibility
- Amazon Mechanical Turk: launched 2005; 500,000+ registered workers by 2011; named after the 1770 chess-playing automaton hiding a human inside
- More than 140 Facebook content moderators in Kenya diagnosed with severe PTSD after reviewing murders, child abuse, and extreme violence for less than $2/hour
- Workers classified as “independent contractors” — shields companies from minimum wage, overtime, health insurance, workers’ compensation
- May 2024: nearly 100 Kenyan data labelers wrote to President Biden describing conditions as “modern-day slavery”
- Corporate subcontracting architecture: Meta → intermediary (Scale AI, Appen, Sama) → local subcontractor → workers; enables plausible deniability at every level
- When African content moderators sued Meta for psychological harm, Meta argued it wasn’t their employer — the local subcontractor was
- The data annotation market projected to reach $6.98–8.2B by 2028–2029
- EU AI Act includes transparency provisions for data workers in AI supply chains
- Kenyan data labelers filed lawsuits against Meta; TikTok moderators in Turkey formed union, faced aggressive union-busting
- 2025 summit: moderators from Kenya, Philippines, India, Latin America united around common demands: direct employment, mental health support, higher wages, right to organize
- The “AI illusion” (ILO term): presenting human labor as machine automation to justify lower prices, reduce accountability, avoid employment regulations
Newsletter Angles
- The ghost work story is the inverse of the Mechanical Turk Pattern: it’s not about a single product being mislabeled — it’s about an entire industry’s business model depending on invisible labor
- The geographic concentration of exploitation is deliberate: companies target countries with high unemployment, educated workforces, and weak labor protections — a race to the bottom that is global and structural
- Content moderation trauma is the AI industry’s deepest ethical scandal. 140+ Facebook moderators in Kenya with PTSD. This is not a side effect — it is the product
- The organizing story is underreported: workers across four continents are building cross-border solidarity. This is labor internationalism for the AI age
Entities Mentioned
- Meta — primary employer of content moderators via subcontracting; sued by African moderators
- OpenAI — former Scale AI client
- Amazon — operator of Mechanical Turk platform; takes 20% commission on all tasks
Concepts Mentioned
- Mechanical Turk Pattern — this source is the definitive account of ghost work as structural AI dependency
- Leverage Erasure Through Automation — the simultaneous dynamic: real automation threatens visible workers while simulated automation exploits invisible workers
- Data Privacy Weaponization — workers reviewing intimate content are themselves harmed; their labor creates AI that processes user data
Quotes
“We do this work at great cost to our health, our lives and our families… for less than $2 per hour.” — Kenyan data labelers, open letter to President Biden
“Our working conditions amount to modern day slavery.” — Kenyan data labelers
“The AI revolution has produced remarkable capabilities. But it’s been built through means that resemble 19th-century exploitation more than 21st-century innovation.”
Notes
This is a synthesis piece from ScienceArray, not primary research. It draws on multiple documented cases, the 2022 Ghost Work book by Mary Gray and Siddharth Suri, and news reporting. Pairs with the AWU/TechEquity US data workers report (which documents similar conditions among US-based workers). The sciencearray piece covers the global/offshore dimension; the AWU report covers the domestic dimension.