Refugee gig workers in Kenya power AI data work for pay they can't predict

In a portable building in Kakuma, a refugee camp near Kenya's northern border with South Sudan, a worker named Grace used ChatGPT to research translations of Christian hymns in languages she does not speak. She did not know the client, the purpose of her work, or whether she would be paid at all. As a refugee unable to legally work in Kenya, she needed any opportunity she could get.

Grace, who requested anonymity due to a nondisclosure agreement, prompted ChatGPT to translate a hymn's English text into an assigned East Asian language. She then used the translation to search for existing versions and to find information about the hymn's translator, author, and Christian denomination. Finally, she fed those details into AOP Connect, a crowdsourced research platform owned by RWS, a British AI enterprise software company.

The Guardian reported on her case as part of a broader investigation into refugee gig workers powering technology services for dwindling pay, with the stark framing that these workers "don't know how much they'll be paid – if AI hasn't already taken their jobs." One worker's summary of the economics driving the sector: "Someone else will do it for less."

Why does this story matter?

The case sits at the intersection of two structural trends in the AI industry: the outsourcing of data and research labor to low-cost gig workers, and the automation pressure that AI models themselves place on those same workers. Refugees in Kenya are legally barred from formal employment, which strips them of bargaining power and makes opaque pay arrangements — or no pay at all — a condition of participation rather than an exception.

The work itself is not incidental to AI. Grace's workflow ran through ChatGPT and ended in a crowdsourced research platform owned by a British enterprise software firm. That means the supply chain for AI-related services can run through a refugee camp, with workers who cannot verify who benefits from their output or what it is worth.

What does the arrangement look like in practice?

According to the reporting:

• Grace worked from a portable building in Kakuma camp, near Kenya's border with South Sudan.
• She was assigned translation research into an East Asian language she does not speak, using ChatGPT to bridge the gap.
• She submitted structured findings — translator, author, Christian denomination — to AOP Connect, a crowdsourced research platform owned by RWS.
• She operated under a nondisclosure agreement and did not know the client or the purpose of the work.
• Payment was uncertain: she did not know whether she would be paid for the task.

Grace's inability to legally work in Kenya is the binding constraint. It converts gig platforms into one of the few available income channels, regardless of terms, and it explains why pay can dwindle — displaced workers can always be replaced by others in the same position, or, increasingly, by the AI systems they help train and feed.

What is at stake going forward?

The story underscores a growing accountability gap in AI labor chains: enterprises consume research and data work delivered through crowdsourced platforms, while the workers at the far end — in this case refugees with no legal right to employment — absorb the risk of nonpayment and the threat of automation. As AI tools like ChatGPT compress the skill required for tasks such as multilingual research, the leverage of the people performing them shrinks further. The Guardian's reporting suggests the question is no longer only what refugee gig workers are paid, but whether the work survives the technology it supports.