📖+10 XP
🎧+10 XP
✅+15 XP
AI moderation misses most African languagesCEFR A1
20 Apr 2026
Adapted from Guest Contributor, Global Voices • CC BY 3.0
Photo by Zulfugar Karimov, Unsplash
Level A1 – BeginnerCEFR A1
2 min
79 words
- AI systems often remove harmful posts from platforms.
- Many African languages are not understood by AI.
- Moderators cannot read several local languages on apps.
- Users sometimes lose content without clear explanation.
- Harmful posts in local languages can stay online.
- This problem hurts creators, journalists and ordinary users.
- Research groups are building datasets for African languages.
- Some projects focus on Hausa, Swahili and Igbo.
- European rules may push platforms to improve moderation.
- Many teams still need more time and support.
Difficult words
- harmful — Causing damage or making people or things hurt
- moderator — Person who checks posts and removes bad onesModerators
- platform — Website or app where people share postsplatforms
- dataset — Collection of text or data for computer studydatasets
- creator — Person who makes videos, posts or online contentcreators
- journalist — Person who writes news stories or reportsjournalists
- support — Help, money, or resources for a team or project
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- Have you ever lost content on an app?
- Do you speak an African language?
- Who is hurt by this problem?
Related articles
8 Dec 2025
22 Dec 2025
2025 aid cuts threaten health and humanitarian services
Large reductions in international aid in 2025 disrupted health and humanitarian services in many low- and middle-income countries. The cuts began with a US suspension of aid and led to the closure of USAID and wider global impacts.
7 Nov 2025
26 Jan 2026
16 Apr 2026
Rethinking 'the Human' in AI
Artist and writer Xonorika Kira argues that centring the human in AI can exclude other forms of knowledge and intelligence. She proposes practical changes like small, consent-based datasets and slower models to support communities and cultural sovereignty.