OpenAI and Meta are teaming up to train AI systems on African languages. This project, which includes French telecom giant Orange SA, aims to fix a major gap in AI. Africa, a continent with over 2,000 languages, has largely been overlooked in AI training.
The project will launch next year and initially focus on two West African languages: Wolof and Pulaar. Together, these languages are spoken by over 22 million people, yet current AI models fail to understand them.
Orange’s goal is to expand its communication tools in Africa and the Middle East, where it operates in 18 markets. But this isn’t just about customer service. The first phase will use data centers in Europe and Africa and public cloud systems to train the AI. Still, the company hasn’t clarified where it will source the training data.
Expanding AI to local languages
Steve Jarrett, Orange’s Chief AI Officer, said the project is only starting with Wolof and Pulaar. By next year, they plan to train AI models for other sub-Saharan languages like Lingala, Swahili, and Bambara. These languages are vital for millions but barely exist in AI datasets.
According to Jarrett, this initiative is a step toward making AI accessible to everyone, even populations that are illiterate or otherwise excluded from digital advances.
OpenAI will give Orange early access to its models. This allows Orange to develop AI-powered tools like voice interactions for customer service. They will also handle data processing and hosting from European data centers.
The models will operate as open-source, free for non-commercial uses such as health and education services. The broader aim? To roll out custom AI solutions for all 18 West African countries where Orange has operations.
For context, Wolof is spoken in Senegal, The Gambia, and southern Mauritania. Pulaar is mainly used in Senegal. Together, these languages represent a significant population often overlooked by mainstream AI systems.
This is a big deal for companies like Orange, which operate in areas where local languages dominate but are completely invisible to tools like OpenAI’s GPT or Meta’s Llama.
Why “sovereign AI” matters
This effort taps into the growing concept of “sovereign AI.” The idea is simple but powerful: regions and countries want control over the core tech infrastructure that powers AI.
By localizing data storage and processing, they ensure their languages, cultures, and histories aren’t erased. For Africa, which has often been sidelined in global tech development, this is a chance to put its linguistic diversity on the map.
Orange’s role here is strategic. The company plans to localize data processing by hosting OpenAI’s models in European centers.
This partnership also gives Orange early access to the latest AI advancements, which they plan to use for applications like voice-powered customer support systems.
From Zero to Web3 Pro: Your 90-Day Career Launch Plan
