Malicious AI Swarms: Researchers in Mali and Ethiopia Anticipate Threat to African Democracy

There is no evidence yet of swarms of autonomous agents conducting large-scale influence campaigns, but their building blocks are making progress. A study examines the risks to African hybrid regimes and proposes a layered response.

Generative AI increases the scale and finesse of disinformation campaigns. An international team is interested in the next step: "malignant AI swarms", multi-agent systems that are coordinated, persistent and adaptable, designed to conduct influence operations. The authors distinguish them from the simple production of AI-assisted content and false profiles controlled from a centre. They state that there is no evidence to date of such self-contained swarms operating on a large scale, but that the capabilities that would make them possible advance.

Specially exposed land

The study focuses on hybrid regimes and conflict-affected states in Africa, where fragile institutions and vulnerable media ecosystems could amplify the effect of these tools: infiltration of communities, building artificial consensus, erosion of trust in authorities. In Mali, several actors are fighting a fragmented information space; in Ethiopia, narrative management is part of a more organized state strategy. Lack of training data in African languages could slow down these campaigns, but also weaken defences. The authors are particularly concerned about hybrid operations combining the automation scale with the knowledge and credibility of local actors.

They propose multi-layered governance, linking technical safeguards, platform accountability, civic institutions and regional coordination, to protect democratic participation, peacebuilding and development.