An artificial intelligence (AI) educator, Ms Zainab Saidu, says most AI tools still struggle to accurately interpret Nigerian languages and cultural contexts, warning that reliance on Western-trained systems could lead to bias and misinformation.
Saidu, founder of the Creativity Enthusiast Network, observed while assessing the performance of AI systems currently in use in Nigeria.
She explained that many tools misinterpret Hausa, Yoruba, Igbo and Nigerian Pidgin, often failing to grasp tone, dialect variations, proverbs and everyday expressions. She added that speech recognition systems also struggle with local accents and background noise, despite gradual improvements driven by local data collection efforts.
ALSO READ SpaceX’s Starlink hits 9m global users
Saidu cautioned that heavy dependence on AI trained mainly on Western datasets poses risks in sensitive sectors such as healthcare, governance and journalism, where misinterpretation could have serious consequences.
According to her, symptoms described in Pidgin or Hausa may be wrongly analysed by AI systems trained largely on Western medical data.
She stressed the need for the development of local datasets that reflect Nigerian linguistic and cultural realities, noting that this would require collaboration among media organisations, technologists, linguists, universities and government.
Saidu acknowledged growing efforts by agencies such as the National Information Technology Development Agency (NITDA) and startups including Awarri and CDIAL AI to develop local language models, but emphasised the importance of ethical guidelines, transparency, journalist training and strong data protection frameworks.
“Safeguards should ensure AI strengthens public trust in journalism without stifling innovation,” she said.
