A comprehensive mapping project has exposed the hidden infrastructure powering artificial intelligence training, tracing connections between major tech companies and the contractors who employ thousands of remote workers to label data in the Global South.
The interactive visualization, created by a Brown University senior researcher, charts 13 data-labeling vendors supplying training data to 141 corporate clients across 31 countries. According to AI Weekly, the project reveals a concentrated labor supply chain with worker concentrations in Kenya, Indonesia, Colombia, and similar developing economies.
Understanding the Data Labeling Economy
Data labeling represents one of the least visible but most critical components of modern AI development. Before large language models and computer vision systems can function effectively, humans must annotate millions of examples: categorizing images, transcribing audio, rating model outputs for quality and safety, and performing countless other classification tasks that establish the ground truth for machine learning systems.
This work has largely shifted to offshore contractors operating in countries with lower labor costs. The mapping project documents how this global division of labor has become the backbone of AI advancement, with major technology companies outsourcing their labeling needs through specialized intermediaries.
What the Data Reveals

The visualization presents several significant findings about AI's supply chain:
- A relatively small number of vendors act as intermediaries between AI companies and workers
- Labor concentration in specific developing regions creates dependency and potentially limits worker bargaining power
- The tasks range from routine image classification to evaluating AI model behavior and safety concerns
- Geographic distribution suggests intentional cost optimization strategies by AI developers
The project makes visible a relationship typically obscured by corporate procurement structures. AI companies rarely employ labelers directly, instead contracting with specialized firms that handle recruitment, management, and quality control. This arrangement creates distance between AI developers and the workers who teach their systems.
Implications for AI Transparency and Labor
The mapping initiative addresses growing concerns about AI transparency and labor practices. As generative AI has accelerated development timelines and computational demands, the data labeling workforce has expanded considerably. Yet documentation of working conditions, compensation, and scale has remained sparse.
The public database enables researchers, journalists, and policymakers to examine corporate connections and potentially trace responsibility for labor practices. It also provides a foundation for future investigation into specific vendor operations and their treatment of workers.
Industry observers have increasingly scrutinized the economics of AI training data, particularly after reports documenting low wages, limited worker protections, and exposure to disturbing content in some labeling operations. This mapping project provides concrete documentation of the networks involved.
Moving Forward
The visualization raises questions that technology companies and regulators will need to address: What labor standards should apply to AI training work? Should transparency about data sourcing and labeling practices become an industry requirement? How can workers in these roles negotiate better terms?
As AI systems become increasingly central to commerce and society, the invisible infrastructure supporting their development continues attracting scrutiny. This mapping project transforms an abstract concern into documented reality, connecting corporate AI ambitions directly to human labor in specific countries and regions.



