- 21/03/2025
- Category: Commentaries
Author : Nadia Sinta Rindiani
Editor : Ayom Mratita Purbandani
Artificial intelligence (AI) is often portrayed as a technology that automates human labor, especially in knowledge-based sectors. However, this oversimplified narrative overlooks a fundamental question: Who actually ‘builds’ AI? Behind the sophistication of self-learning algorithms lies the essential role of invisible human workers, such as data annotators who label images, freelancers who refine datasets, and content moderators who filter out harmful material. While automation has replaced many knowledge-based tasks, these ‘ghost workers’ remain significant for training and maintaining AI systems. Ironically, their contributions often go unrecognized and unprotected, unlike those of formal workers. Trapped in a “Human-as-a-Service” model, they carry out essential yet temporary tasks, receiving low wages with no job stability and security guarantees.
The ghost workers behind AI: How invisible workers keep AI running
While AI is capable of automating various knowledge-based tasks, such as data analysis, natural language processing, and even content creation, it still relies significantly on invisible human labor, including data curation, model evaluation, and other preparatory tasks[1]. One of the key aspects of AI development, yet overlooked, is data labeling, which involves tagging and annotating text, images, and videos to enable machine learning systems to recognize patterns and make predictions. This form of labor is indispensable to the functionality of technologies such as ChatGPT, autonomous cars, and large language models (LLMs). Interestingly, this type of work is often referred to as ‘ghost worker,’ because despite playing a crucial role in ensuring AI training data quality, these workers remain largely unrecognized[2]. Furthermore, despite the centrality of their tasks to the AI systems, many of these workers are often rewarded with low wages and little recognition[3,4]. A contemporary analysis highlights that the compensation for such labor is alarmingly low, with many of these workers receiving hourly wages that are significantly below the standard living wage, often amounting to less than $3 per hour[5]. The heavy reliance on ghost workers reveals a major paradox in AI development. On the one hand, AI is automating more and more jobs, but, on the other hand, the demand for low-wage project-based labor is increasing[6,7]. This phenomenon is inseparable from the complexity of the supply chain in the AI industry. Far from being the product of incorporeal intelligence, AI systems rely on the transnational web of labor, infrastructure, and resources. Many companies rely on outsourcing labor from different regions to perform tasks such as data annotation and information processing. Geographical and organizational factors play an important role in the asymmetries of power and visibility, where workers are often employed in conditions that are far from ideal[2].
The majority of ghost workers come from Global South countries such as the Philippines, Kenya, India, Pakistan, Venezuela, and Colombia[8]. They are exploited with low wages and little protection. According to the 2022 Contract Worker Disparity Project report, despite the work of outsourced workers being essential to keeping tech companies in business, there is a clear and significant gap between them and full-time employees[9]. On average, subcontracted workers earn only 75 cents for every dollar earned by permanent staff. In addition, they often do not receive basic employment benefits such as paid leave, health insurance, job security, or the prospect of career advancement[9]. These workers generate huge economic value for companies but are not rewarded accordingly. This reveals a global division of labor in which essential but menial tasks are devalued. As a result, global social and economic inequality is exacerbated[10].
In addition to wage disparities the work of these ‘ghost workers’ is often monotonous and stressful[11]. Without developing skills or economic mobility, they remain trapped in jobs that offer no opportunity to improve their quality of life, and limit their ability to grow, adapt, and improve their long-term prospects. Meanwhile, big tech in developed countries continue to rely on cheap labor without having to provide benefits or social security. The resulting profits are not fairly distributed, further reinforcing inequality in the global labor market[12, 13]. In other words, the efficiencies that technology promises often come at a high price, at the expense of workers’ rights and security, deepening social and economic inequalities in the world of work.
The human as a service model: Exploitation and precarity
The “Human-as-a-Service” (HaaS) work model further exemplifies the intensification of precarious labor conditions in the digital age. This model turns human labor into a “service” that can be activated or deactivated as needed, so workers are treated like interchangeable assets rather than individuals with agency and rights. While gig work platforms capitalize on the flexibility of this model as a benefit, ironically, such rhetoric opens up opportunities for exploitation. Workers are underpaid, overworked, and denied basic benefits such as health insurance or pensions[3,14,15]. Consequently, stable forms of employment have shifted to short-term tasks without protection.
Workers in this sector, often recruited through platforms such as Amazon Mechanical Turk, Appen, and Scale AI, are typically compensated on per-task completion rather than hourly wage. This type of remuneration results in their incomes being unstable and economically vulnerable[3]. This reliance on low-paid digital workers unprotected social security to sustain AI ecosystems reflects a global trend that increasingly marginalized many of the important jobs behind the digital ecosystem[16]. In addition, this technology-driven economy is able to suppress workers’ bargaining power regarding wages and weaken working conditions. The impact is not only on individual job security but also threatens the social fabric by undermining workers’ rights and collective bargaining power. As automation replaces traditional jobs, there is a risk of losing the important achievements of the labor movement in fighting for justice and dignity of work[17].
The rise of AI presents a paradox wherein technological progress coexists with labor exploitation. AI automates many knowledge-based jobs but remains dependent on cheap labor, especially in countries of the Global South. Ghost workers handling labeling data often receive low wages without social protection, while AI companies enjoy huge profits. This shows that human labor is indispensable but remains marginalized.
To ensure that the benefits of technology are more equitable, there is a pressing need for labor policy reforms that guarantee workers’ rights and the application of ethical standards in AI development[18]. Recent research has proposed a “charter of rights” for ghost workers to protect them in the evolving gig economy[19]. With this approach, worker welfare can go hand in hand with technological innovation and can create a more inclusive and equitable future.
Endnotes
1 Muldoon, J., Cant, C., Wu, B., & Graham, M. (2024). A typology of artificial intelligence data work. Big Data & Society, 11(1), 20539517241232632.
2 Tubaro, P., Casilli, A., & Coville, M. (2020). The trainer, the verifier, the imitator: Three ways in which human platform workers support artificial intelligence. Big Data & Society, 7(1), 2053951720919776. https://doi.org/10.1177/2053951720919776
3 Duffy, B. (2020). Algorithmic precarity in cultural work. Communication and the Public, 5(3-4), 103-107. https://doi.org/10.1177/2057047320959855
4 Gandini, A. (2019). Labour process theory and the gig economy. Human Relations, 72(6), 1039-1056.
5 Muldoon, J., Cant, C., Graham, M. et al. (2025). The poverty of ethical AI: impact sourcing and AI supply chains. AI & Soc 40, 529–543. https://doi.org/10.1007/s00146-023-01824-9
6 Melián-González, S., & Bulchand-Gidumal, J. (2021). What type of labor lies behind the on-demand economy? New research based on workers’ data. Journal of Management & Organization, 27(5), 850-866.
7 Sulumov, S. (2023). Transformation of the vocational training system in the conditions of digitalization of the labor market. SHS Web of Conferences, 172, 01031. https://doi.org/10.1051/shsconf/202317201031
8 Pogrebna, G. (2024). AI is a multi-billion dollar industry – it’s underpinned by an invisible and exploited workforce. The Conversation. Available at: https://theconversation.com/ai-is-a-multi-billion-dollar-industry-its-underpinned-by-an-invisible-and-exploited-workforce-240568 [Accessed 20 March 2025].
9 Bracy, C. (2023). The ghost workforce the tech industry doesn’t want you to think about. Stanford Social Innovation Review. Available at: https://ssir.org/articles/entry/the_ghost_workforce_the_tech_industry_doesnt_want_you_to_think_about [Accessed 13 April 2025]
10 Konstantis, K. (2024). The main challenges of AI ethics: Historical contextualization, black-boxing, social biases, labor invisibility. In Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (Vol. 7, No. 2, pp. 23-25). https://doi.org/10.1609/aies.v7i2.31899
11 Plantin, J. C. (2021). The data archive as factory: Alienation and resistance of data processors. Big Data & Society, 8(1), 20539517211007510. https://doi.org/10.1177/20539517211007510
12 Ignatieva, I., Saraeva, O., Zedgenizova, I., & Zvezdina, A. (2023). The impact of the digital economy on employment and labour market transformations: The main trends. International Journal of Innovative Research and Scientific Studies, 6(3), 507-513. https://doi.org/10.53894/ijirss.v6i3.1554
13 Kaun, A., Velkova, J., Laaksonen, S., Delfanti, A., Logsdon, A., Stiernstedt, F., … & Ruckenstein, M. (2020). In the shadows of the digital economy: the ghost work of infrastructural labor. AoIR Selected Papers of Internet Research. https://doi.org/10.5210/spir.v2020i0.11131
14 Insebayeva, S., & Beyssembayev, S. (2023). Digital platform employment in Kazakhstan: Can new technologies solve old problems in the labor market? International Labor and Working-Class History, 103, 62-80. https://doi.org/10.1017/s0147547923000200
15 Kneese, T., & Palm, M. (2020). Brick-and-platform: Listing labor in the digital vintage economy. Social Media + Society, 6(3). https://doi.org/10.1177/2056305120933299
16 Munn, L. (2024). Digital labor, platforms, and AI. In Werthner, H., et al. Introduction to Digital Humanism. Springer, Cham. https://doi.org/10.1007/978-3-031-45304-5_35
17 Acemoglu, D., & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3-30. https://www.aeaweb.org/articles?id=10.1257/jep.33.2.3
18 Fischer, L., Wunderlich, N., & Baskerville, R. (2023). Artificial intelligence and digital work: The sociotechnical reversal. Hawaii International Conference on System Sciences. https://hdl.handle.net/10125/102655
19 Friedland, J., Balkin, D., & Montealegre, R. (2020). A ghost workers’ bill of rights: How to establish a fair and safe gig work platform. California Management Review Insights.
