Ghana Summit Calls for Clear AI Accountability in African Elections Amid Misinformation Surge

A recent Electoral Integrity Summit in Accra, Ghana, highlighted the urgent need for African nations to establish clear accountability frameworks for AI-generated content that could harm electoral processes. Professor Henry Kwasi Prempeh from the Ghana Center for Democratic Development emphasized that the debate should move beyond simply adopting AI to defining who is responsible when these technologies cause damage, particularly in the context of political campaigns.

The summit's concerns are grounded in recent real-world examples. Anthropic identified a Kenyan actor using its Claude models to create mass political content for the 2027 elections, while Nigeria's Defence Headquarters disowned an AI-generated deepfake video of a military chief. Zambian police also issued warnings about AI-driven misinformation targeting officials. These incidents underscore the immediate threat AI poses to democratic integrity across the continent.

Prempeh argued that the issues of political finance, technology, and civic space are interconnected. Undisclosed campaign funding can be used to purchase large-scale AI-generated content, and restricted civic spaces hinder the ability of journalists and observers to expose such abuses. This creates an asymmetry where the cost of creating damaging misinformation has fallen dramatically, while detection and correction remain expensive.

The recommendations from the summit included calls for greater transparency in political finance, robust safeguards for AI and information integrity, and enhanced protection for media and citizen observers. Speakers also stressed that technology alone cannot build political trust and that cooperation among states, civil society, and democratic institutions is essential to defend electoral integrity throughout the entire election cycle.

An underlying technical challenge, not directly addressed by the summit, is the difficulty of AI misinformation detection in low-resource African languages. Research indicates that translation quality significantly impacts detection performance, meaning systems may struggle with languages like Hausa or Twi, where political narratives often originate. This technical gap, coupled with Meta's phasing out of third-party fact-checking in Africa, complicates the task of establishing accountability for AI-driven electoral harm.