Artificial intelligence is often judged by the performance of its models. Discussions typically focus on accuracy, bias mitigation, explainability, or regulatory compliance. Much less attention is given to the governance of the data that make these systems possible.

This session argues that AI cannot be considered trustworthy if the underlying data are collected, governed, and used in ways that reproduce unequal power relations. Better algorithms built on unjust data practices simply automate existing inequalities at greater speed and scale.

Drawing on research from African digital ecosystems, including Digital Public Infrastructure, civic technology, FemTech, and digital labour platforms, the presentation introduces data justice as a practical framework for evaluating AI systems. It examines questions of consent, ownership, participation, representation, accountability, and community agency across the AI lifecycle.

Participants will learn how data justice differs from conventional AI ethics, why it should be integrated into AI design and governance from the outset, and how organisations can identify data governance risks before they become AI risks. The session concludes with a practical framework that researchers, developers, policymakers, and technology leaders can apply when designing or evaluating AI systems