Open Veins of Algorithmic Auditing: Why AI Assessment Lags Behind Its Deployment in the Global South

Massive deployment-governance gap in the Global South: AI systems are being deployed across Latin America, Sub-Saharan Africa, and Asia Pacific at rates matching or exceeding the Global North, yet fewer than twenty published audits of deployed systems exist across the entire region over a decade, despite hundreds of documented public-sector algorithms and multibillion-dollar national AI investments.

Four recurring technical failures in Global South AI systems: The paper identifies persistent patterns including proxy targets that prioritize predictability over validity, performance claims that fail under prevalence analysis, populations being scored by models that never encountered similar data during training, and structural bias that remains even after removing protected attributes—suggesting systemic rather than isolated problems.

The audit gap is fundamentally a funding problem, not a capacity problem: The authors argue that the scarcity of AI evaluation in the Global South stems from lack of funding for audits rather than lack of technical expertise, since "capacity follows funded demand" and auditing resources have not been directed to these regions.

Existing audit infrastructure is inadequate for Global South contexts: Current evaluation frameworks developed in the Global North don't account for the distinct regulatory, linguistic, and data conditions of the Global South, making standard approaches insufficient for assessing AI systems deployed in these regions.

Critical systems affecting vulnerable populations lack oversight: The paper documents audits of high-stakes systems like child-welfare risk models and public-employment matching algorithms that directly impact vulnerable populations, yet most remain unreleased and unexamined by independent evaluators.