Social media platforms operating across Africa rely heavily on AI moderation systems that are almost entirely blind to the continent's linguistic diversity, with research showing that only 42 African languages appear in any meaningful way in major AI models and just four — Amharic, Swahili, Afrikaans, and Malagasy — are handled with any consistency. The gap produces two distinct failures: harmful content in languages the system cannot parse goes undetected, while legitimate posts are removed without explanation, as seen in Kenya where TikTok's automated systems took down more than 592,000 videos in a single quarter of 2025. The burden falls disproportionately on African creators, journalists, and ordinary users, even as the data-labelling work that trains these systems is performed largely by workers in Kenya, Nigeria, and other African countries — a stark asymmetry that may also put platforms in breach of EU regulations, including the Digital Services Act and the AI Act, which require non-discriminatory systems and meaningful explanations for moderation decisions.