AI’s unsung heroes: The underpaid workers powering global AI development

AI development depends on human-labeled data - millions of images, texts, and audio clips meticulously categorized to train machine learning models. This work, known as data annotation and verification, is outsourced through global online platforms like Amazon Mechanical Turk, Clickworker, and Microworkers.

Artificial intelligence (AI) is often portrayed as an independent and autonomous technology, revolutionizing industries and replacing human labor. However, hidden beneath the surface of AI innovation is a massive yet invisible workforce - data workers who train, refine, and validate AI models. These workers are responsible for labeling data, verifying AI outputs, and even performing tasks AI cannot yet complete. While AI is often depicted as replacing jobs, it simultaneously creates new forms of labor that are frequently precarious, low-paid, and outsourced to workers in developing economies.

A recent study, "The Digital Labour of Artificial Intelligence in Latin America: A Comparison of Argentina, Brazil, and Venezuela" by Paola Tubaro, Antonio A. Casilli, Mariana Fernández Massi, Julieta Longo, Juana Torres Cierpe, and Matheus Viana Braz, sheds light on the realities of AI data work in Latin America. Conducted through surveys and interviews with over 900 workers across Argentina, Brazil, and Venezuela, the research highlights economic hardship, labor informality, and inequality as defining characteristics of AI-driven digital labor in the region. The study uncovers how Latin American workers contribute to the global AI supply chain while facing low wages, job insecurity, and exclusion from legal protections.

The rise of AI data work in Latin America

AI development depends on human-labeled data - millions of images, texts, and audio clips meticulously categorized to train machine learning models. This work, known as data annotation and verification, is outsourced through global online platforms like Amazon Mechanical Turk, Clickworker, and Microworkers. Tech companies, primarily in North America and Europe, contract these platforms to manage the labor-intensive process of AI model refinement, while workers in lower-income countries complete the tasks.

In Latin America, economic instability and high unemployment have driven many workers to turn to AI data work as a source of income. The study found that while most data workers in the region have university degrees, their earning potential remains extremely low. Many workers perform repetitive and menial tasks for as little as $1 per hour, despite their high levels of education and technical skills. The appeal of this work lies in its flexibility and payment in U.S. dollars, particularly in countries like Venezuela, where local currencies have drastically depreciated.

However, the precarity of AI data work is evident. Unlike formal employment, workers receive no job security, social benefits, or career growth opportunities. AI companies treat data workers as disposable contractors, relying on a constant influx of new workers to sustain the labor-intensive process of AI development. This system mirrors traditional patterns of economic exploitation, where wealthier nations outsource labor to lower-income regions while extracting high-value digital products in return.

Economic hardship and informal labor: The reality of AI workers

The study reveals stark differences in how AI data work is experienced across Argentina, Brazil, and Venezuela, shaped by each country's unique economic and social context.

In Argentina, inflation and economic instability have led many young, educated workers to seek side incomes through AI data work. While most Argentine workers use data work as a supplementary income source, earning an average of $83 per month, few rely on it as a full-time job. The country’s complex currency exchange system further complicates payments, as many workers resort to cryptocurrencies and informal financial channels to access their earnings.

In contrast, Brazilian data workers are more likely to be women, unemployed individuals, and informal workers who turn to AI platforms out of necessity. Nearly 40% of Brazilian AI data workers are women, many of whom balance care responsibilities and informal jobs with data work. The informal labor market in Brazil is deeply rooted in historical inequalities, and AI work follows the same trend, offering low wages and limited social protections. Many workers earn an average of $112 per month, often working unpredictable hours due to the global nature of AI data tasks, which are assigned based on client time zones in North America and Europe.

Venezuela presents the most extreme case, where AI data work is a primary income source for 75% of workers surveyed. The country’s economic collapse and political instability have led Venezuelan workers to flood international data work platforms, accepting some of the lowest wages globally. Many earn around $70 per month, but due to Venezuela’s deteriorating economy, even these low earnings provide a lifeline for survival. The study highlights how Venezuelan data workers invest in digital payment methods and use virtual private networks (VPNs) to access better-paying AI tasks - strategies that reflect their desperate efforts to remain competitive in a highly exploitative digital labor market.

AI data work: A new form of digital inequality

Despite their contributions to AI development, Latin American data workers remain invisible within the global AI ecosystem. Their labor is essential for training AI systems used in self-driving cars, voice assistants, and facial recognition, yet they are excluded from the wealth and progress generated by these technologies. The study argues that AI data work reinforces digital colonialism, where developing countries provide cheap digital labor while developed nations reap the financial rewards.

One of the most concerning findings is that AI data work offers little to no career advancement. Unlike traditional outsourcing industries, which may provide opportunities for skills development and job progression, AI data work is fragmented, anonymous, and highly replaceable. Workers are assigned microtasks that require no long-term expertise, ensuring that their roles remain easily automated or reassigned to lower-cost workers elsewhere. This leads to a cycle of digital exploitation, where AI companies benefit from a vast, disposable workforce with minimal investment in worker well-being.

Moreover, AI data work amplifies existing gender and economic inequalities. In Brazil, the study found that women in informal jobs are disproportionately drawn to AI platforms, using them to fill income gaps left by precarious employment conditions. In Venezuela, workers with higher education and technical skills are forced into low-paying AI tasks instead of contributing to local innovation and economic growth. These trends highlight how AI development is deepening social divides rather than creating new opportunities for economic mobility.

Towards a fairer AI economy: Policy recommendations

The study underscores the urgent need for better regulations to protect AI data workers and ensure fair compensation. Governments, international organizations, and tech companies must recognize that AI development is not fully automated - it relies on human labor, which must be valued and protected.

One key recommendation is to enforce fair wages and ethical labor standards for AI data workers. Digital labor platforms should be transparent about pay rates and working conditions, preventing companies from hiding behind outsourced contracts to exploit workers. Additionally, policymakers must advocate for worker protections, ensuring that AI data work includes social benefits, legal rights, and pathways for professional development.

Another crucial step is to increase awareness about AI labor exploitation. AI companies rarely acknowledge the human workforce behind their models, allowing them to market AI as fully autonomous while ignoring the hidden digital workforce supporting its development. Raising public awareness about the human cost of AI production can pressure companies to adopt ethical AI labor practices and invest in more inclusive, fairer AI supply chains.

Ultimately, the study highlights that AI is not built in isolation - it is fueled by thousands of underpaid, invisible workers across Latin America and other developing regions. As AI continues to expand, it is crucial to ensure that its benefits are shared equitably and that the global digital workforce receives the recognition and compensation they deserve.