Who legitimises the AI algorithm? Leadership, volatility and the governance of algorithmic authority

Purpose This paper reconceptualizes ethical AI governance as a leadership problem of legitimacy production rather than technical compliance. Existing frameworks presume stable infrastructure, coherent institutions, and baseline trust − assumptions that frequently fail in volatile environments. Drawing on Africa as a theory-generating extreme context, this paper reframes ethical governance as a legitimacy infrastructure required for governable AI deployment, addressing the question of who legitimizes the AI algorithm. Design/methodology/approach The paper develops an analytically grounded conceptual framework supported by illustrative vignettes. It specifies volatility as three-dimensional: infrastructural, institutional, and socio-political, showing how each generates distinctive governance breakdowns. Building on legitimacy theory and algorithmic accountability scholarship, the paper derives a Sensing–Stabilizing–Legitimizing (SSL) leadership capability model and three falsifiable propositions that explain how organizations sustain contestability and accountability amid instability. Findings In times of volatility, ethical AI governance succeeds only when leaders institutionalize legitimacy infrastructure rather than relying on principles alone. Infrastructural volatility produces exclusion-by-fragility; institutional volatility generates compliance theater; socio-political volatility amplifies legitimacy shocks. The SSL capability chain is decisive for sustaining governable algorithmic authority. Practical implications Leaders should establish clear decision rights, escalation pathways, and recourse mechanisms calibrated to volatility conditions. Organizations should build redundancy into monitoring, prioritize safeguards for vulnerable groups, and treat impact assessments as living governance instruments rather than one-time compliance deliverables. Originality/value The paper makes an integrative-conceptual contribution by theorizing volatility as an explicit governance condition and positioning ethical AI governance as a strategic leadership capability rather than a delegated technical task. It positions Africa as an extreme context that reveals hidden assumptions in dominant governance models and generates transferable insights for rising global volatility.