•  American tech giants have outsourced AI training to countries such as Kenya, where critics say they are exploiting cheap labor and weak protections
• Workers face short-term contracts, low pay and emotional trauma from tagging graphic content
• ‘This is a continuation of slavery and colonialism. It’s our labor and our raw materials that continue to fuel these revolutions,’ says tech rights advocate Angela Chukunzira

NAIROBI, Kenya 

American tech giants are tapping into Kenya’s desperate youth labor market to fuel the development of artificial intelligence, but civil society groups warn the promise of opportunity masks a deeper crisis of exploitation.

In what activists are calling “AI sweatshops,” thousands of young Kenyans are being hired to train artificial intelligence systems for global firms like US social media company Meta, OpenAI, Microsoft, Google and Amazon.

Their tasks include sorting, labeling and annotating images, videos and text – the invisible backbone of AI’s rapid expansion, shaping everything from chatbots to surveillance systems.

But behind this technological boom lies a troubling reality, says Angela Chukunzira, a tech rights advocate and lead at Project Ether, who has been closely tracking how global AI development intersects with labor rights in Kenya.

Speaking to Anadolu, she warned that the contradiction between opportunity and exploitation defines much of Kenya’s AI labor. “This is a case of both. It is an opportunity for Kenyans to engage in building AI tools, but it comes with a lot of exploitation … The problem comes in when these people are underpaid, overworked.”

Hired through subcontractors under often opaque agreements, many workers find themselves bound to short-term contracts, sometimes lasting only days or weeks, without job security or benefits.

The nature of the work is monotonous, highly repetitive and can be emotionally draining.

Workers click through endless volumes of data, tagging objects, transcribing dialogue and drawing digital boundaries around figures in images so algorithms can “learn” to see, speak and behave, Chukunzira explained.  

Human cogs in the AI machine

This work, known in the tech industry as human-in-the-loop (HITL), is critical to refining machine learning systems. Yet despite its centrality, it is deeply undervalued, outsourced to regions like East Africa where high unemployment and economic hardship have created fertile ground for digital exploitation.

Chukunzira raises concerns over the lack of transparency in data labeling tasks, which leaves workers unsure of the purpose behind their assignments.

“You have no full disclosure information of what this data is going to be used for. Why am I doing this? Sometimes, the invisible, hidden agendas can be a big problem. Ethically, morally, it’s wrong.”

In Kenya, where youth unemployment hovers above 67%, millions of young people are desperate for work. For many, this kind of digital labor, however precarious, appears to offer a way out.

Brian Mwende, 22, a former university student, left his studies due to financial pressure and turned to this industry for income.

“The money isn’t enough, but it’s something. Sometimes you work for two weeks and then wait another month for new tasks,” he said.

Another worker, Melvin Ogada, a recent graduate of Jomo Kenyatta University of Agriculture and Technology, described spending up to eight hours a day annotating violent and graphic images for a content moderation tool.

The pay was meager, the hours long and the psychological toll heavy.

“You don’t know if your contract will be renewed,” Ogada said. “Some of us are let go without notice, and we don’t get proper mental health support, especially when dealing with traumatic content.”

Faith Nyawira, 29, an IT student and single mother, joined an AI labeling project after hearing it paid more than her teaching job.

“I ended up tagging images of violence and car crash victims,” she said. “No one told me that’s what the work would involve, but I needed the money.”

“They told us it was part of the future, and we feel proud to contribute,” she said. “But we’re left guessing if our work makes any real difference or if we’re just cheap labor from the village.”

Mental health, privacy and labor rights are among the key concerns raised by activists and researchers tracking the outsourcing of AI work.

Chukunzira said that what sets digital labor apart from factory work is its invisibility, which undermines solidarity and erases the human effort behind AI.

“These are people who are hidden and invisible. AI is actually made to look like this shiny, automated thing. Many people do not know that it's actually people behind it.”

Kenya’s rise as a digital outsourcing hub, driven by its young, English-speaking population and expanding fiber optic infrastructure, has been celebrated by both government officials and industry leaders as a sign of progress.

President William Ruto has touted digital jobs as a cornerstone of the country’s economic transformation agenda, but critics argue that without strong legal safeguards, transparent contracts and access to grievance mechanisms, the workers behind billion-dollar AI systems remain vulnerable to abuse.

Meta and OpenAI did not respond to multiple requests from Anadolu for a comment on the issue.  

‘Moral and ethical failure’

Chukunzira pointed out that digital workers in Kenya lack the basic protections enjoyed by factory workers, leaving them vulnerable to instability as the digital gold rush sweeps across the globe.

“Somebody is working on a project for three weeks, and after that they’ll not have a job for like four months. Unlike the traditional factory floor, there’s no regular income. This instability is much more in the case of new data workers,” she said.

This digital divide is not unique to Kenya. Across the Global South, a growing underclass of digital workers is training algorithms that will shape the future, even though they themselves may never fully share in its benefits.

In places like the Philippines, India and Uganda, similar concerns are being raised about exploitative tech labor practices marketed as empowerment.

This trend, Chukunzira says, reflects a deeper historical injustice: “This is a continuation of slavery and colonialism. It’s our labor and our raw materials that continue to fuel these revolutions.”

She criticized major tech companies for neglecting mental health and failing to provide long-term protections to those doing difficult annotation work.

“There is a moral and ethical failure from Big Tech companies in terms of the workforce that they’re employing to do this kind of work,” she said.