The hidden cost of AI: Africa’s invisible workforce and digital servitude

Each day, Mercy logs in from a windowless cubicle in Nairobi. Her task is to watch. For hours, she reviews gruesome images, listens to hate-filled audio, and flags content that AI systems cannot yet comprehend. She is not a tech executive or engineer. She is a data worker, paid just over $1 an hour to clean the internet for the world—while the trauma remains lodged in her memory.
Mercy’s story is not unique. It’s the rule. Across Africa, thousands of workers like her silently have been powering theartificial intelligence revolution. Their keystrokes train chatbots. Their eyes filter content for safety. Their voices fuel speech recognition. But their names remain hidden. Their stories were ignored. Their rights denied.
The glossy rhetoric of AI—efficiency, progress, human enhancement—masks a system that increasingly relies on cheap, expendable labor from the Global South. While companies in the Global North tout ethical AI and “responsible innovation,” African workers are being discarded, underpaid, and silenced. This isn’t innovation. It’s exploitation with a new interface.
Tech giants like OpenAI, Meta, and Google depend on vast volumes of labeled data to train their models. This labor is offshored to firms like Sama Source, which source workers from Kenya, Nigeria, and beyond to annotate text, classify images, and moderate toxic content. These tasks are essential for AI development—but treated as disposable.
Since 2023, investigations have documented chilling realities. Workers exposed to graphic violence and sexual abuse content are left without trauma counseling. Some develop PTSD. Others are fired without cause or bound by contracts that forbid them from speaking out. The emotional toll is invisible to end-users—but hardwired into the AI systems they rely on.
This isn’t just about poor working conditions. It’s about power. Africa, once colonized for its land and minerals, is now being mined for data and digital labor. This is digital colonialism: a system where value is extracted, exported, and monetized elsewhere, while the costs—economic, emotional, and ethical—are borne by Africans.
These companies thrive in a governance vacuum. Fewer than half of the African nations have enforceable data protection laws. Even fewer have meaningful labor protections for digital workers. The African Union’s Malabo Convention remains ratified by only 15 countries.
The Writer is a Global AI and Climate Policy Strategist, AI Researcher at UNEP and a Technology Journalist