From mineral extraction to manufacturing, delivery to waste, there are hundreds of millions of laborers³⁴⁹. And when you consider that AI technologies such as algorithms are trained from huge amounts of user data³⁵⁰, we also have to consider the billions of people online who’s data is necessary for these tools to work.³⁵¹ People whose data is mined are not compensated and oftentimes don’t even realize their contributions, which have even included private messages over LinkedIn³⁵². “[P]eople are not just users or citizens; they are data points, commodities and, ultimately, labourers whose activity generates profit but not recognition.”³⁵³ To most people, these millions or billions of workers — a significant number of whom earn low wages and deal with poor labor conditions — are essentially invisible and unacknowledged. Moreover, the constant demands of efficiency and economics come with consequences. When people are expected to work like machines, it leaves little space for care and comfort. Having everything available at the touch of a screen may seem nice – but what is the human cost?

There is no single report documenting workers in all areas of technology, from mining and extraction of minerals, to production and assembly of hardware to development of software, infrastructures needed to make the internet function, distribution and delivery, and e-waste management. However reports in each of those areas already point to the total being hundreds-of-millions of workers in just a few parts of the lifecycle of tech – a 2023 report estimated up to 435 million online gig workers³⁵⁴; one 2011 report noted a 100,000 person factory in China whose sole product was the Apple iPad,³⁵⁵ and another 2017 report noted 411,000 electronics industry workers in Vietnam alone where 80% were women working on assembly lines³⁵⁶.

The sheer mass of people needed to create and maintain the internet and internet-connected technologies cannot be overstated, and the conditions of those laborers is important to look into. There is very little documented about working conditions, proportionate to the numbers of laborers, but what is known does more than simply raise eyebrows.

AI, when functioning well, requires a lot of work to train – datasets that are built and codified by people³⁵⁷. Digital labor platforms centralize and distribute work at the click of a button – popular platforms include Clickworker, Appen, and Amazon's Mechanical Turk. These websites make it possible for annotation tasks to be quickly distributed 24/7 to a huge amount of workers around the world – most tasks only earning a few cents per assignment. The data is compiled into valuable datasets used to train AI models.³⁵⁸

The name “Mechanical Turk” refers to an automated chess-playing device in the 18th century, where challengers would fall for the illusion that they were competing against a machine, but in fact an actual person was hiding inside the machine, making the decisions and doing the real work³⁵⁹. The concept is reminiscent of the story “The Wonderful Wizard of Oz” when the illusion of ‘The Great and Powerful Oz’ is accidentally revealed to be an ordinary man hiding behind the mystical persona. ”[...] Toto jumped away [...] and tipped over the screen that stood in a corner. As it fell with a crash they looked that way, and the next moment all of them were filled with wonder. For they saw, standing in just the spot the screen had hidden, a little old man, with a bald head and a wrinkled face, who seemed to be as much surprised as they were.“³⁶⁰ In the same vein, AI has been given the mythology that it is like-magic, but behind the scenes, ordinary people are doing the work. “These workers are kept hidden for a reason. They say if slaughterhouses had glass walls, the world would stop eating meat. And if tech companies were to reveal what they make these digital workers do, day in and day out, perhaps the world would stop using their platforms.”³⁶¹

Some reports claim that up to a third of the tasks on digital labor platforms, such as Amazon’s Mechanical Turk, are unpaid³⁶² because the ‘requesters’ "can simply refuse payment to workers if they are dissatisfied with the job - and it is not worth the annotator's time chasing a few unpaid cents.”³⁶³ Doing click work and data annotation might be a simple way to make a quick dime, but it is not sustainable. According to researcher Julian Posada: “social reproduction when labor is commodified is unsustainable because it depends on a normative embeddedness that negatively affects the livelihoods of workers, creates dependencies, and contributes little to the long term development of local communities.”³⁶⁴ There is an assumption that click work is easy, but that is not necessarily the case. Researcher Milagros Miceli expressed that labelling satellite imagery was “extremely challenging [...] strenuous on the eyes and hands.” One worker expressed concern to Miceli, “I hope this isn’t for weapons.”³⁶⁵

From virtual information ‘factories’ to physical concrete-building factories, productivity is maximized, and not uncommonly to the detriment of the workers. Kate Crawford, AI researcher warned: “There’s the concern that robots are replacing us, but actually what’s happening is that we’re being treated more and more like robots.”³⁶⁶

Extensive commuting, pressure to reach daily targets, psychological stress, dangerous work-spaces, low salaries, short contracts, and monitoring and surveillance, are just a few of the inhumane conditions facing factory workers.³⁶⁷

One report from 2011 about a tech hardware-production factory company in China said: “Talking and stretching are forbidden on the assembly line, and clocking in five minutes late may result in the loss of half a day’s wages. Bathroom use is limited to 10 minutes, which is strictly enforced by an electronic key card.”³⁶⁸ A decade later, in 2021, it was reported that one factory worker in China died after working 380 hours per month³⁶⁹– this would equate to two or two-and-a-half-times the legal maximum working hours in Germany. It isn’t just China – a study of Vietnam’s female factory workers described how they worked “alternating day and night shifts for periods of 4 days; standing for the entire 9 – 12-hour shift.”³⁷⁰ The UN reported further about the factory worker conditions in Vietnam: “Some of the surveyed women reported adverse health consequences associated with unhealthy working conditions including miscarriages, extreme fatigue and fainting.”³⁷¹

Beyond factory workers, warehouse workers and delivery drivers at centers distributing and shipping to shops and directly to consumers also deal with notoriously inhumane working conditions. One undercover investigation in the UK found that Amazon warehouse workers and delivery drivers were urinating in trash cans, and defecating in cars to meet their strict quotas.³⁷² Another study on Amazon warehouses in the UK found that ambulances were called 600 times in three years.³⁷³ Amnesty International stated that Amazon warehouse worker conditions in Saudi Arabia were equivalent to those of human trafficking. “In Saudi Arabia, the [Amazon warehouse] workers were mostly housed, for months, in dirty and overcrowded accommodation, sometimes infested with bed bugs. They were put to work in Amazon warehouses, but the contractors often withheld part of their salaries and/or food allowances without explanation, and underpaid overtime.”³⁷⁴ At one point, Amazon attempted to patent a cage for warehouse workers.³⁷⁵

In recent years, a number of exposes looking into the working conditions of gig workers³⁷⁶ and content moderators have come to light. In 2023, “Silicone Savanna” looked into social media’s content moderator working conditions in Kenya – a decade-old booming industry, employing approximately 100,000 people worldwide through third-party outsourcing companies, and estimated to be valued at $40 billion USD by 2032. “[Despite] what some of Silicon Valley’s other biggest names tell us, artificial intelligence systems are insufficient moderators. So it falls on real people to do the work.” Workers are exposed to psychologically harmful content from hate speech to child exploitation to mass murders for less than two dollars per hour, and are expected to maintain a high quota each shift. “The word evil is not equal to what we saw,” said one content moderator working for a third-party company in Kenya that was contracted by Facebook. Another remarked of her four years on the job, “I don’t even know how I survived. It was the worst time of my life."³⁷⁷ Click workers and data labellers who are making next to nothing³⁷⁸, conduct their work using mandatory automated tools experience surveillance³⁷⁹ and punishment when they deviate from the expected fast pace³⁸⁰.

These stories are in abundance, and for the billions of people who use social media, the working conditions are inconceivable. A content moderator aptly noted “I feel like it’s modern slavery, like neo-colonialism.”³⁸¹ Neo-colonialism typically refers to one state taking power over another, and this is a very modern capitalistic example of neo-colonialism, where the companies for which the content moderators are moderating are based in the United States and bringing the billions of dollars of net worth back to that country. The net worth is not finding its way to Kenya, or anywhere else the content moderation is happening. “What emerges is a new kind of dependency, where digital infrastructures replicate the old colonial logics: extraction without redistribution, occupation without direct governance, dominance without accountability.”³⁸² A term has even emerged to refer to this phenomenon: digital colonialism.³⁸³

Although labor rights are demanded as a basic requirement of this kind of set-up, rights alone would not resolve this exploitative system which uses the labor of people in countries of the so-called ‘Global Majority’ to ultimately and significantly benefit the companies based across Europe and headquartered in the United States, which is, by far, the wealthiest nation on Earth. Because the work is divided into such small tasks, which can be done by almost anyone, the knowledge about the system as a whole cannot be decentralized. That's why people are easily replaceable with cheaper, faster and ‘non problematic’ workers.

"[T]asks are distributed across a global workforce, with the most stable, well-paid and desirable jobs located in key cities in the US, and the most precarious, low-paid and dangerous work exported to workers in peripheral locations in the Global South [continuing] colonial patterns of Western countries leveraging their economic dominance and growing rich off extracting minerals and labour from peripheral territories."³⁸⁴

These are power structures that run deep. One example noted that India supplies a quarter of the world’s web-based work, but only receives 3% of the global revenue from online labor platforms.³⁸⁵

Labor unions are formalized groups formed by workers in a common sector, field, or company, in order to unify and advocate for better working conditions, higher wages, etc. Using methods like collective bargaining, labor unions make the demands stronger as a collective, rather than as individuals behind closed doors. Each country and context has different laws and attitudes toward labor unions. In the United States, for example, although the federal law technically prohibits employers from taking action against employees who are forming a union, in practice companies may dismiss those workers anyway, since the penalties are low.³⁸⁶ While it may seem like there is a lot of buzz around labor unions, actually only 10% of American workers belong to one³⁸⁷ – although 60 million workers in the US would like to join one but cannot³⁸⁸. But the tide is turning.

Around the world are hopeful stories of laborers fighting for their rights. Mercy Mutemi is a litigation lawyer in Kenya who is at the forefront of worker protection suits against Big Tech companies and their third-party subcontractor companies. Mutemi sets the issue at hand in clear terms: “I think when you give people work for a period of time and those people can’t work again because their mental health is destroyed, that doesn’t look like lifting people out of poverty to me. That looks like entrenching the problem further”.³⁸⁹

WhatsApp content moderators in Texas sued Meta for proper compensation for the difficult work they had been subjected to³⁹⁰. In Kenya, one whistleblower, Daniel Motaung, sued Meta for exploitation and union busting while another active case includes 185 content moderators from across Africa who were working for a Kenya-based Meta contractor, Sama, who are seeking $1.6 billion in compensation.³⁹¹ In 2023, an Indian startup garnered attention for compensating AI language trainers at a rate that relative to the local market was considered fair.³⁹²

After working for Amazon for years, during the COVID-19 pandemic, Chris Smalls realized that worker protections were not in place at his warehouse and the situation was becoming ever more dire.³⁹³ Colleagues were coming in to work sick, and Smalls was getting worried about his and his colleagues’ wellbeing. “My job description says ‘have a high school diploma and lift fifty pounds’. It doesn’t say ‘risk my life working during a pandemic’.”³⁹⁴ He started organizing a labor union. This was risky, as he could have gotten fired for doing so. It was far from easy, and took a lot of grassroots organizing and face-to-face conversations with colleagues. Smalls started first in his own warehouse, talking to his co-workers on breaks and on the way to work. “We would speak everywhere — in the staircase, in the elevator, in the parking lot, in front of the building, on the bus, at St George Ferry Terminal”.³⁹⁵ In 2022, after a great deal of engaged and active participation, he and his colleagues were the first Amazon workers to be recognized by the US National Labor Relations Board.³⁹⁶ Later, the movement grew until the Amazon Labor Union³⁹⁷ was formed across warehouses in the country.

While much of the data, anecdotes, and footage of ‘invisible labor’ is bleak, this essay ends on a hopeful note, best encapsulated in a quote by Smalls: “The young generation is paying attention now. The uprising of all these different movements that’s being led by younger adults. You know, Amazon being a trillion dollar company, all the odds were against us. But we’ve proven that no amount of money in the world can amount to the power of people.”³⁹⁸

³⁴⁹ Hussein, Fatima. “Online gig work is growing rapidly, but workers lack job protections, a World Bank report says.” Associated Press, 2023.

³⁵⁰ Paul, Katie; et al. “Inside Big Tech's underground race to buy AI training data.” Reuters, 2024.

³⁵¹ Porter, Eduardo. “Your Data Is Crucial to a Robotic Age. Shouldn’t You Be Paid for It?” The New York Times, 2018.

³⁵² da Silva, João. “LinkedIn accused of using private messages to train AI.” BBC, 2025.

³⁵³ Delclós, Carlos. “Countering Digital Colonialism.” CCCB LAB, 2025.

³⁵⁴ Datta, Namita; et al. “Publication: Working Without Borders: The Promise and Peril of Online Gig Work.” World Bank, 2023.

³⁵⁵ Cheng, Sophia. “The deadly labor behind our phones, laptops and consumer gadgets.” Colorlines, 2011.

³⁵⁶ IPEN. “Stories of women workers in Vietnam’s electronics industry.” 2017.

³⁵⁷ Dzieza, Josh. “AI Is a Lot of Work.” The Verge, 2023.

³⁵⁸ Pontin, David. “Artificial Intelligence, With Help From the Humans.” The New York Times, 2007.

³⁵⁹ Hitlin, Paul. “Research in the Crowdsourcing Age, a Case Study: What is Mechanical Turk?” Pew Research Center, 2016.

³⁶⁰ Baum, L. Frank. “The Wonderful Wizard of Oz.” Originally published in 1900 by George M. Hill Company. Re-published under an open license on Project Gutenberg in 1993.

³⁶¹ Cockerell, Isobel. “In Kenya’s slums, they’re doing our digital dirty work.” Coda Story, 2025.

³⁶² Gray, Mary L.; et al. “Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass.” Harper Business, 2019.

³⁶³ Cant, Callum; et al. “Feeding the Machine: The hidden human labour powering A.I.” Bloomsbury, 2024.

³⁶⁴ Posada, Julian. “Embedded Reproduction in Platform Data Work.” Information, Communication & Society, 2022.

³⁶⁵ Miceli, Milagros. ““I hope this isn’t for weapons.” How Syrian data workers train AI.” Untold Mag, 2024.

³⁶⁶ Crawford, Kate. “Digital Talk ‘Living with Artificial Intelligence’ with Kate Crawford and Guy Hoffman.” (at 21:30) Vitra Design Museum, 2023.

³⁶⁷ Cant, Callum; et al. “Feeding the Machine: The hidden human labour powering A.I.” Bloomsbury, 2024.

³⁶⁸ Cheng, Sophia. “The deadly labor behind our phones, laptops and consumer gadgets.” Colorlines, 2011.

³⁶⁹ Feng, Emily. “Scandals in China bring calls for more regulation over tech companies.” NPR, 2021.

³⁷⁰ IPEN. “Stories of women workers in Vietnam’s electronics industry.” 2017.

³⁷¹ OHCHR. “Vietnam: UN experts concerned by threats against factory workers and labour activists.” 2018.

³⁷² Godlewski, Nina. “Amazon working conditions: urinating in trash cans, shamed to work injured, list of employee complaints.” Newsweek, 2018.

³⁷³ Wilding, Mark. “Ambulances were called to Amazon warehouses 600 times in three years.” VICE, 2018.

³⁷⁴ “Saudi Arabia: Migrants workers who toiled in Amazon warehouses were deceived and exploited.” Amnesty International, 2023.

³⁷⁵ Picchi, Aimee. “Amazon's patent for caging workers was "bad" idea, exec admits.” CBS News, 2018.

³⁷⁶ Rest of World. “The Graying Gig Workers series.”

³⁷⁷ Hellerstein, Erica. “Silicon Savanna: The workers taking on Africa’s digital sweatshops.” Coda Story, 2023.

³⁷⁸ Rowe, Niahm. “Millions of Workers Are Training AI Models for Pennies.” Wired, 2023.

³⁷⁹ Privacy International. “Humans in the AI loop: the data labelers behind some of the most powerful LLMs' training datasets.” 2024.

³⁸⁰ Williams, Adrienne. “The Exploited Labor Behind Artificial Intelligence.” Noema, 2022.

³⁸¹ Perrigo, Billy. “Inside Facebook's African Sweatshop.” Time, 2022.

³⁸² Delclós, Carlos. “Countering Digital Colonialism.” CCCB LAB, 2025.

³⁸³ Okinyi, Mophat. “African Digital Colonialism is the New Face of Worker Exploitation.” ICTworks, 2025.

³⁸⁴ Cant, Callum; et al. “Feeding the Machine: The hidden human labour powering A.I.” Bloomsbury, 2024.

³⁸⁵ Pearson, Tamara. “How companies in the Global North use online gig work to exploit Global South workers.” The Real News Network, 2023.

³⁸⁶ Hertel-Fernandez, Alexander. “Labor law makes it too hard to start unions. Workers deserve a bigger voice.” CNN Business, 2020.

³⁸⁷ Podhorzer, Michal. “The Paradox of the American Labor Movement.” The Atlantic, 2024.

³⁸⁸ Shierholz, Heidi; et al. “Workers want unions, but the latest data point to obstacles in their path.” Economic Policy Institute, 2024.

³⁸⁹ Cockerell, Isobell. “Legendary Kenyan lawyer takes on Meta and Chat GPT.” Coda Story, 2024.

³⁹⁰ Perrigo, Billy. “WhatsApp Content Moderators Review the Worst Material on the Internet. Now They’re Alleging Pay Discrimination.” Time, 2021.

³⁹¹ Musambi, Evelyne. “Facebook loses jurisdiction appeal in Kenyan court paving the way for moderators' case to proceed.” The Independent, 2024.

³⁹² Perrigo, Billy. “The Workers Behind AI Rarely See Its Rewards. This Indian Startup Wants to Fix That.” Time, 2023.

³⁹³ Yohannes, Aron. “Chris Smalls, Amazon Labor Union Founder, On Digital Activism.” The Mozilla Blog, 2024.

³⁹⁴ Day, Meagan. “'Amazon Is a Breeding Ground'.” Jacobin, 2020.

³⁹⁵ Blanc, Eric; et al. “The Workers Behind Amazon’s Historic First Union Explain How They Did It.” Jacobin, 2022.

³⁹⁶ O’Brien, Sara Ashley. “Amazon workers at New York warehouse vote to form company’s first US union.” CNN Business, 2022.

³⁹⁷ Amazon Labor Union.

³⁹⁸ Rise25. “How Chris Smalls took on Amazon and made history.” Mozilla (published on YouTube), 2024.