Towards an African Agenda for AI Safety

Artificial intelligence is reshaping economies, politics, and societies worldwide, and Africa is no exception. While AI promises benefits for health, education, agriculture, and economic growth, this report highlights how its risks are especially acute on the continent given fragile institutions, limited infrastructure, and geopolitical dependencies. Key risks identified include: Malicious use: deepfakes and AI-powered disinformation threaten elections and social cohesion; AI-enabled surveillance raises concerns for human rights; cybercrime and online gender-based violence are growing. Malfunction: reliability issues from models trained on Western data, bias in AI tools, and risks to critical sectors like healthcare and education. Systemic risks: labour disruptions in sectors such as business process outsourcing, worsening environmental pressures from compute and data centres, and escalating e-waste. Despite the African Union’s Continental AI Strategy, national-level efforts remain fragmented. Only a handful of states (Kenya, Ghana, Morocco, Rwanda, South Africa) show meaningful activity on AI safety, and no African country currently hosts a dedicated AI Safety Institute. The paper advances a five-pillar action plan to build a robust African AI safety ecosystem: Grounding AI governance in human rights to protect the most vulnerable. Establishing an African AI Safety Institute to lead research, testing, and policy innovation. Expanding public AI literacy through education and targeted campaigns. Building early warning systems with multilingual benchmarks for at least 25 African languages. Convening an annual AU-level AI Safety & Security Forum to align policy, coordinate responses, and amplify Africa’s voice in global governance. By embedding African perspectives into global AI safety frameworks, the agenda seeks to safeguard communities, assert technological sovereignty, and shape a fairer, more resilient AI future.