When we hear the term “Artificial Intelligence,” the images that typically come to mind are highly paid engineers in Silicon Valley, sleek data centers, and advanced autonomous systems ready to usher humanity into a utopian future[cite: 7]. Major tech leaders speak of a coming era where AI cures disease, manages energy grids, and creates unparalleled prosperity. However, the TRT World documentary In The Belly of AI peels back this glossy narrative to reveal a disturbing reality: the so-called “artificial” intelligence is built on the sweat, trauma, and exploitation of millions of invisible human workers, largely located in the Global South, and comes at a steep environmental cost[cite: 7].

The Illusion of Autonomy

Milos Michelli, a sociologist and computer scientist, highlights a fundamental misconception: we define AI as “intelligent systems” that conduct tasks without human intervention, but this is a myth[cite: 7]. “AI is just a bunch of work, it’s just a bunch of resources, it’s just a bunch of labor coming together to produce an outcome,” Michelli explains[cite: 7]. The narrative of complete machine autonomy intentionally conceals the massive human labor force required to make these systems function[cite: 7].

To understand why this labor is necessary, we must look at how AI models are trained[cite: 7]. Whether it is OpenAI, Meta, DeepMind, or xAI, every AI company requires billions of data points to teach their algorithms[cite: 7]. For a driverless car to identify a pedestrian, or a chatbot to provide a safe response, the raw data must first be manually sorted, labeled, and annotated by a human being[cite: 7].

According to a World Bank report, there are currently between 150 million and 430 million data workers globally[cite: 7]. The vast majority of these workers are located in the Global South—in countries like India, Kenya, Venezuela, and Lebanon[cite: 7].

The Global South as a Digital Sweatshop

The outsourcing of data work to the Global South is not accidental; it is a calculated strategy by big tech to exploit vulnerable populations in crisis-stricken economies where labor rights are weak and wages are abysmal[cite: 7]. As one expert notes, companies target desperate populations and “pay hardly enough for workers to make ends meet… to survive to eat that day, but not tomorrow.”[cite: 7]

Arnab Tas, a 29-year-old data annotator in Kolkata, India, explains his daily routine of classifying images for AI models[cite: 7]. But while classifying an image of a penguin or an active fire may seem innocuous, the reality for many data workers is deeply traumatic[cite: 7].

The Trauma of Content Moderation

To ensure that AI chatbots like ChatGPT do not produce harmful, violent, or sexually explicit content, they must first be taught what that content looks like[cite: 7]. To do this, tech giants send massive datasets containing the darkest, most disturbing corners of the internet—and the dark web—to workers in countries like Kenya for manual review[cite: 7].

“Our job is to make sure that we clean this work for you, for you to enjoy AI in these western countries,” explains a Kenyan data worker[cite: 7]. Workers describe reviewing content that is almost exclusively toxic, including graphic violence, murder, rape, and child exploitation[cite: 7].

Fine Makira, a former AI data worker, recounts the severe psychological toll of watching corpses and abuse videos day after day[cite: 7]. “I used to have nightmares… I developed anxiety, I could not go out… it kind of led me to depression,” Makira shares[cite: 7]. Many workers develop severe PTSD, insomnia, and anxiety as a direct result of their employment[cite: 7].

Furthermore, these workers are often forced to sign strict non-disclosure agreements, threatening them with jail time if they speak out about the nature of their work[cite: 7]. When workers attempt to unionize to demand better conditions or mental health support, they are swiftly terminated, a practice that violates core international labor conventions[cite: 7].

The Environmental Price Tag

Beyond the human toll, the physical infrastructure of AI is devastating to the environment[cite: 7]. “I don’t think there is a way of having technologies without harming nature,” Michelli states[cite: 7].

The industry intentionally obscures its physical footprint, using terms like “the cloud” to hide the reality of massive, energy-intensive data centers[cite: 7]. In order to function, AI requires gargantuan quantities of metals and rare earth minerals—copper, cobalt, lithium, and tungsten—drawn from extensive mining operations[cite: 7]. Furthermore, manufacturing AI chips and cooling massive data centers consumes vast amounts of water and electricity, directly contributing to the climate crisis[cite: 7].

The “TESCREAL” Ideology

How do Silicon Valley leaders justify this immense human suffering and environmental degradation? The answer lies in a cluster of ideologies prevalent in the tech sector, known by the acronym TESCREAL (Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, and Long-termism)[cite: 7].

Dr. Emil P. Torres, a philosopher and historian, explains that “Long-termism” is built on a techno-futuristic vision where humanity builds an Artificial General Intelligence (AGI) that will allow a “posthuman” species to colonize the universe[cite: 7]. Because the tech elite believe they are building a utopian paradise for future millennia, they view the present-day exploitation of the Global South and the destruction of the environment as a minor, morally justifiable sacrifice[cite: 7].

The industry’s big bosses have quietly proclaimed themselves the creators of a paradise, utilizing an anti-democratic, elitist vision that harms the planet and exploits the vulnerable[cite: 7]. As the documentary concludes, it leaves us with a chilling question: “After the workers of the Global South, after the mines dug, the rivers dried up, the atmosphere heated up, who do you think will be the future victims in the belly of AI?”[cite: 7]