How Hyper-Datafication Impacts the Sustainability Costs in Frontier AI

A comprehensive study of 550,000 datasets from Hugging Face reveals that the AI industry's rapid scaling of data collection—termed 'hyper-datafication'—disproportionately shifts environmental, labor, and social costs to the Global South and precarious workers. The research identifies critical sustainability challenges in frontier AI development and proposes the Data PROOFS framework to mitigate representational harms, carbon footprint, and labor exploitation.

The transition to hyper-datafication represents a fundamental shift in how frontier AI models are developed. Rather than primarily training on existing data, technology corporations now actively generate and curate data at scale specifically for model development. This intensification of data infrastructure creates measurable environmental and social costs that the field has largely overlooked in discussions about AI sustainability.

The environmental implications are substantial. Large-scale data centers required for storage and processing consume significant energy, generating carbon emissions distributed unevenly across regions. The research reveals that data infrastructure development disproportionately burdens the Global South, where many data centers operate but whose populations receive minimal benefits from the resulting AI systems. This geographic inequality compounds existing global disparities in resource allocation and environmental responsibility.

Labor dimensions amplify these concerns. Direct interviews with data workers in Kenya expose precarious employment conditions, inadequate compensation, and psychological harm from exposure to graphic content during data annotation and curation tasks. These roles, essential to frontier AI development, remain systematically undervalued and underprotected.

For the AI industry, this analysis suggests mounting pressure toward greater transparency and sustainability standards. The Data PROOFS recommendations—spanning provenance, resource awareness, ownership, openness, frugality, and standards—indicate emerging expectations for corporate accountability. Investors and developers should anticipate regulatory frameworks addressing data ethics and carbon footprint reporting. The research establishes baseline metrics for measuring these costs, enabling stakeholders to track progress toward more sustainable AI development practices.

• →Hyper-datafication shifts environmental burdens, labor risks, and representational harms systematically toward Global South nations and precarious workers.
• →Analysis of 550,000 datasets reveals measurable storage-related energy consumption and carbon footprint impacts previously overlooked in AI sustainability discourse.
• →Data workers in Kenya face precarious employment conditions, inadequate compensation, and psychological harm from exposure to graphic content.
• →The Data PROOFS framework proposes six principles—provenance, resource awareness, ownership, openness, frugality, and standards—to mitigate data-related sustainability costs.
• →Global disparity in data center infrastructure concentrates environmental and economic benefits in developed nations while concentrating costs in the Global South.