Driven Down: Amazon delivery drivers and workplace technologies (www.dair-institute.org)

A new critique of Amazon’s delivery technologies argues that the harms we attribute to “AI” often rest on labor exploitation at scale: route-optimization, real-time tracking, predictive scheduling and performance-score systems compress human work into data, intensify pace, and enable punitive management decisions. The piece highlights how these socio-technical systems rely on drivers for labeled data, edge-case recovery, and constant feedback — yet drivers bear the physical risks, precarious employment conditions, and disciplinary surveillance while having little say over system design or remediation. For the AI/ML community this reframes technical debates about fairness, transparency and accountability as fundamentally political and transnational labor issues. Rather than only improving model explainability or audit tools, the authors call for centering worker organizing and collective bargaining as an ethical priority: empowering transnational worker voice can change data collection practices, evaluation metrics, incentive designs, and deployment policies that currently prioritize operational efficiency over safety and dignity. The implication is clear for practitioners and policymakers — build socio-technical governance that recognizes workers as co-creators of AI systems, and shift ethics interventions from technical fixes to structural workplace protections.