In Kakuma refugee camp, near Kenya's border with South Sudan, more than 300,000 residents depend almost entirely on humanitarian aid. Fewer than half of households have any income, and those that do earn less than $50 per month. Into this desperation stepped the global tech outsourcing industry, promoted by UN agencies and companies alike as a 'win-win' — refugees get income, AI companies get cheap labor. The reporting from the Guardian shows conclusively which side won. The Solidarity Initiative for Refugees trained more than 2,000 refugees in digital skills since 2016, connecting over half with remote tech work. For several years, workers on platforms like Remotasks — a Scale AI subsidiary — earned 500 to 1,000 Kenyan shillings ($3 to $6) per day doing data annotation, transcription, and translation. That pay was never generous, but it was guaranteed and valued in a camp with almost no economic alternatives. Then Remotasks shuttered its Kenya operations in 2024, and the microwork economy began its collapse. What replaced guaranteed wages is structurally worse. British AI enterprise software company RWS operates AOP Connect, a crowdsourced research platform where refugees now compete for 'discretionary rewards' rather than earn wages. A hymn-translation project advertised $9,000 in total rewards, split among an unknown number of contributors based on opaque quality assessments. Workers sign NDAs, don't know their clients, and receive no transparent explanation of how pay is determined. RWS told the Guardian that reward structures are communicated 'transparently upfront' and that its platform is 'not AI-enabled or used for AI-model training work.' The trajectory is unmistakable. The International Trade Centre estimates that transcription, data entry, translation, and web research jobs have declined roughly 50% since 2022. Oxford Internet Institute research corroborates this, showing popular AI tool adoption coinciding with measurable drops in exactly the job categories promoted as paths to refugee self-sufficiency. Na'amal, a nonprofit linking refugees with employers, noticed clients requesting skilled specialists rather than basic digital literacy workers. One refugee, Merci Biamungu, trained in 3D modeling, described his final project: teams of five competing to design a 3D potato model, with the winning group splitting about $116. The AI tools he helped train now build 3D images faster, eliminating the need for workers like him. The human costs go beyond wages. Refugee data annotators in Kakuma reported being tasked with labeling images and videos of weapons, gruesome bodily wounds, and pornography. Workers fleeing wars experienced PTSD symptoms after annotating violent material. The Bureau of Investigative Journalism found that Somali speakers in Kakuma had done transcription work likely used by the US military — work refugees told the bureau they didn't consent to. The opacity of supply chains meant workers often had no idea what they were building or for whom. Joan Kinyua, founding president of the Data Labelers Association, calls these workers the tech sector's 'invisible architects' — foundational to the industry but hidden by opaque supply chains. She experienced the erosion firsthand: wages declining, mistakes penalized harshly, availability demanded around the clock. 'You can't afford to stay away from your computer,' she said. By 2020, she was sleeping only a few hours a night to catch gigs as they appeared. The structural logic is clear. Tech companies accessed an extraordinarily vulnerable labor pool with no bargaining power, no legal right to work in Kenya, no alternative employment, and no ability to enforce contracts across jurisdictions. They extracted the training data and annotations needed to build the AI systems that now eliminate the need for those same workers. The UN and humanitarian sector provided the recruitment infrastructure and legitimacy. The workers got a few years of subsistence-level income, psychological harm from content moderation, and then obsolescence. Bahana Hydrogene, who built the SIR's digital programs, put it plainly: 'They can't get transparent answers from these companies.'
