AI moderation misses most African languagesCEFR A2
20 Apr 2026
Adapted from Guest Contributor, Global Voices • CC BY 3.0
Photo by Zulfugar Karimov, Unsplash
AI systems that remove harmful content often do not understand most African languages. This matters because these systems decide what stays online and what is taken down for many users across the continent.
A 2025 study found only 42 African languages appear in major language models, and just four are handled with consistency: Amharic, Swahili, Afrikaans and Malagasy. Because English data dominates, systems can remove posts without clear explanation or fail to spot harmful content in local languages.
Researchers and groups in Africa are building datasets and models, and new rules in Europe and national strategies may push platforms to improve moderation.
Difficult words
- harmful — causing harm, damage, or danger to people
- content — information like text, images, or videos online
- moderation — the process of checking and removing online posts
- dataset — a collection of data used by computer systemsdatasets
- model — a computer program that learns from datamodels, language models
- platform — websites or apps where people share contentplatforms
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- Have you ever seen an online post removed without a clear explanation? What happened?
- How could better language support help people who use social platforms in Africa?
- What rules or actions could make platforms improve moderation for local languages?
Related articles
Sportellino: a multilingual chatbot for migrants in Italy
A multilingual chatbot called Sportellino launched to help migrants in Italy find public services and practical guidance. It is free, anonymous and available via messaging apps; its information is based on official sources and experts.
Inequality and Pandemics: Why Science Alone Is Not Enough
Matthew M. Kavanagh says science can detect viruses and make vaccines fast, but rising inequality makes pandemics worse. He proposes debt relief, shared technology, regional manufacturing and stronger social support to stop future crises.
Data centers and mining could raise US power costs by 2030
A new study warns that growing electricity use by data centers and cryptocurrency mining could raise power prices and CO2 emissions in the United States by 2030. Researchers modelled hourly power needs across 26 regions to find likely impacts.