Algorithmic Accountability and Data Protection Impact Assessments: Reimagining AI Governance in Africa

The rapid adoption of AI systems across Africa has exposed limitations within existing data protection accountability frameworks. While Data Protection Impact Assessments (DPIAs) have become an established regulatory mechanism under contemporary African privacy regimes, their design remains largely influenced by traditional models of personal data processing, characterised by identifiable data flows, predictable purposes, and relatively stable risks. This article argues that such frameworks are insufficient for governing adaptive AI systems whose risks evolve across their lifecycle. It proposes a reconceptualisation of DPIAs as dynamic algorithmic accountability mechanisms capable of addressing AI-specific concerns including data provenance, model opacity, bias, automated decision-making, and continuous monitoring. Through comparative analysis of African data protection frameworks alongside developments in the European Union, Canada, Singapore, and international AI governance standards, the article develops an AI-enhanced DPIA model suited to African regulatory realities. It argues that Africa need not await entirely new AI legislation but can strengthen AI governance by modernising existing accountability instruments.