Data colonialism

The Moral Cost of Cheap Data: How Emerging Markets Power the AI Boom

By Jack Zhai & Shan Qi

In a cramped room in Caracas, the air conditioner hums against the humid heat. A young woman sits in front of a laptop, her eyes scanning a kaleidoscope of street scenes—thousands of images of traffic lights, pedestrians, and storefronts. She isn't exploring the world. She is training a self-driving car to navigate San Francisco.

For this work, she might earn $200 a month. In the United States, that is a grocery bill. In Venezuela, where the minimum wage has collapsed to the equivalent of $5.40, it is a lifeline.

This is the stark reality of the "intelligence" economy. We are often told that AI is a post-material revolution—a triumph of code over matter. In truth, it is a resource extraction industry. But instead of drilling for oil or mining for lithium, Silicon Valley is mining human cognition. And just like the resource booms of the past, the raw material is found in the Global South, while the value is refined and sold in the North.

Data Colonialism Without the Slogans

Academics call this "data colonialism". It sounds like a slogan, but the economics are brutally simple.

Western AI models are voracious learners. They need massive amounts of "ground truth"—human-verified data to learn from. Hiring an engineer in Palo Alto to label images costs $100 an hour. Hiring a university graduate in Kenya or Colombia to do it costs less than $2.

This arbitrage creates a new kind of trade route. Data flows North, labeled and structured, to build trillion-dollar models like GPT-4 or Gemini. Pennies flow South, usually via crypto rails like USDT, to keep the annotators alive.

It is a symbiotic relationship, but a predatory one. The "Grey" economy of the Global South provides the friction that Western labor laws have removed. There are no benefits, no job security, and no career ladder. There is only the task queue. When the queue is empty, the income stops.

The Trap of the "Digital Gig"

For the workers, this industry is a trap disguised as an opportunity.

In places like Venezuela or rural Brazil, the formal economy has often failed. Hyperinflation, corruption, or simple lack of investment means traditional jobs don’t exist. Into this void step platforms like Appen, Remotasks, and DignifAI.

They offer a compelling pitch: work from home, earn in hard currency (dollars or stablecoins), and be part of the tech future. For a "digital nomad" in Buenos Aires, earning in USDT is a savvy hedge against the peso. But for a data labeler in a slum, it is a survival mechanism.

The MIT Technology Review found that many of these workers are effectively paid piece rates that amount to pennies per task. If they make a mistake—mislabeling a pedestrian, missing a violent phrase—their "accuracy score" drops. If it drops too low, they are banned. It is "algorithmic management" at its coldest: you aren't fired by a boss; you are deactivated by a script.

The Psychological Toll

The cost isn't just financial; it's psychological. A significant portion of "data labeling" is actually content moderation. To teach an AI what "hate speech" or "violence" looks like, a human has to see it first.

Workers in Kenya have reported spending hours filtering videos of beheadings, child abuse, and sexual violence to keep American social media feeds clean. They absorb the toxicity of the internet so that users in the West don't have to.

This labor is invisible by design. The companies building these models go to great lengths to hide the human wiring. They want you to believe the machine is magic. Admitting that the magic trick relies on traumatized workers in Nairobi ruins the illusion.

Policy vs. Reality

So, what is the solution? Western activists call for "fair trade data" and living wages. International organizations like the OECD publish reports on the "platform economy," urging better standards.

But the cynical reality of the Great Game suggests a different outcome. If labor costs in Kenya rise, the platforms won’t necessarily pay more. They will just move. They will find the next currency collapse, the next desperate population—perhaps in Myanmar or rural Bangladesh.

The AI industry is nomadic. It follows the path of least resistance and lowest cost. Until the Global South builds its own "sovereign AI"—models trained on local data, by local workers, for local profit—this dynamic will not change.

For now, the "price of intelligence" remains artificially low, subsidized by the invisible, precarious labor of millions who are training the very systems that may one day replace them.

Digital black markets

Myanmar’s Digital Black Markets: The New Frontier of Global Finance

By Jack Zhai & Shan Qi

On the banks of the Moei River, dividing Thailand from Myanmar, stands a city that technically shouldn’t exist. Shwe Kokko is a $15 billion sprawl of neon-lit casinos, luxury villas, and dormitory blocks rising out of the jungle. It has its own power grid...