South African PhD Research Develops Multilingual AI for Misinformation Detection in isiZulu and Sepedi

A North-West University PhD study by Dr. Seani Rananga has created a groundbreaking multilingual AI framework designed to identify misinformation in English, isiZulu, and Sepedi. This research addresses a critical gap in existing AI tools, which predominantly focus on English, leaving speakers of indigenous African languages vulnerable to the rapid spread of false information. The framework was piloted using COVID-19 misinformation, demonstrating its potential for broader application in areas like elections and public health emergencies.

The study's key technical findings highlight two crucial points: first, that properly designed multilingual AI can effectively detect misinformation across different languages; and second, that the accuracy of machine translation significantly impacts the AI's performance in low-resource languages. This emphasizes the importance of high-quality translation as a foundational element for developing robust AI tools for African language content, rather than an auxiliary feature.

Dr. Rananga's work has garnered international recognition, including a Google PhD Fellowship and the Best Poster Award at the 2025 Deep Learning IndabaX South Africa. This recognition underscores the global importance of her research in fostering more inclusive and reliable AI systems. Her framework offers a vital tool for governments, fact-checking organizations, and social media platforms in South Africa to combat misinformation and ensure greater access to credible information for indigenous language speakers.

The research aligns with South Africa's national policy objectives, particularly the government's plans to establish a national fact-checking platform and implement AI content disclosure rules. It also responds to calls for AI systems that can competently reason in African languages to ensure neutrality in public discourse. Dr. Rananga plans to expand her research to cover more South African languages, address misinformation during elections, and tackle related harms like deepfakes and hate speech, ultimately aiming to build trustworthy multilingual AI systems for various sectors across Africa.