Generative artificial intelligence is reshaping how information is produced, accessed, and circulated, while enabling disinformation campaigns of increasing scale and sophistication. There is currently no clear evidence that fully autonomous AI swarms conduct influence operations at scale, but their enabling capabilities are advancing. We define malicious AI swarms as coordinated, persistent, and adaptive multi-agent systems designed for influence operations, distinguishing them from AI-assisted content production and centrally managed synthetic personas. We examine their implications for hybrid regimes and conflict-affected states in Africa, where institutional constraints and fragile media environments may heighten vulnerability. Drawing on Mali and Ethiopia, we consider how automated influence could infiltrate communities, fabricate consensus, and erode trust in governance and development. The cases illustrate different configurations of state and non-state influence: competing actors in Mali's fragmented information environment, and more organized state-led strategies of narrative management in Ethiopia. African-language and training-data asymmetries may constrain influence capabilities while weakening defensive responses. Hybrid human-AI operations could combine automated scale and adaptation with local knowledge and credibility. This forward-looking risk analysis develops a scenario of increasingly accessible AI-driven coordination, rather than claiming that autonomous swarms are already operating at scale in Africa. We propose a layered governance approach linking technical safeguards to platform accountability, civic institutions, and regional coordination to protect democratic participation, peacebuilding, and development.

翻译：暂无翻译