Throughout the decades, the potential of artificial intelligence has offered little access to individuals with knowledge of African languages. Although there are more than 2,000 languages spoken in the continent, the developments in AI have not considered them much. That story is starting to change now. In an innovative new project, French telecom giant Orange has recently revealed it will use OpenAI to task its newest models to collaborate with African languages, thus, finally providing digital access to an otherwise largely technology-poor, linguistic powerhouse.
The dilemma: Why African languages fall behind in AI
Artificial Intelligence survives on data. The more orderly and digital the data in a language is the better models can be trained on it. Regrettably, African languages do not have huge digital imprints as our languages such as English or Mandarin. Cornell University researchers and publications such as Nature have highlighted how a lack of available data, poor funding and limited computational resources have disregarded African languages in developing AI systems.
Closing that gap is what Orange is working on.
The strategic partnering of Orange and OpenAI
Orange entered into an agreement with OpenAI last year, and it has access to the pre-release AI models. These are not any old models - these models include the whisper speech model that is used by Open AI and offers multilingual transcription features and more recently open-weight models with easy access training parameters.
It is a game changer. As opposed to closed-source models, open-weight ones can be fine-tuned on data other than what was used to create them in the first place: it is possible in theory to adapt Orange to do a specific linguistic operation using its own collection of African languages and thus avoid starting from zero.
This is a partnership that is growing in 2025. Orange will use the latest models of OpenAI in its African markets where it already has an operation in 18 African countries. The vision? To calibrate AI models with authentic pieces of the localized languages of the particular areas in mind and then offer it free of charge to governments and authorities.
What Is at Stake
The influence of Orange OpenAI African languages project extends to translation tools or chatbots. It is all about inclusion. It is all about making sure the next generation of AI apps are able to learn and take into account the opinions of millions of people whose languages have traditionally been left out of the online world.
By localizing the models and putting them to use, Orange will build an environment in which the African languages will become first-class citizens of the AI revolution. This is not only a betterment of skills but also a declaration of fairness.
In the words of Steve Jarrett, who is the Chief AI Officer of Orange:
WE consider this project as a model of how AI can contribute to fixing the digital divide worldwide: by partnering up with local startups and communities, Orange and OpenAI aim to spawn a local ecosystem in which African languages are first citizens of the AI world.
The Bigger Picture Building an AI Ecosystem in Africa
This is not in isolation. Orange is not just creating tools, but is enabling an ecosystem. Plan incorporates:
- Integration of the models with the real world applications by partnering with local startups and developers to create applications with the models.
- Openly sharing fine-tuned models with local officials, enabling governments to create solutions to problems in education, healthcare and other services to the community.
- This supports the concept of encouraging regional data collection and ownership, where the African communities are given power to determine the use of their linguistic data.
It is the opposite of a standard top to bottom technology in which worldwide corporations set the conditions. Its Orange OpenAI African languages is rather symbolic of a more partnership-wise, comprehensive orientation to sustainable change within the scope of the system of digital transformation.
An Example to Other Markets?
The implications of this initiative are not confined to the African continent, as much as the particular initiative itself is. The same problem occurs in other areas where there are languages underrepresented South Asia, Latin America, Indigenous people. The tools and lessons conceived in the process of Orange and OpenAI alliance could become a model, which, in turn, could be copied in other regions of the world.
The ingredients? The right to powerful AI, localized data, and community engagement and intended service to the greater good rather than self-gain.
What Is Next?
Orange has already commenced by using Whisper to turn-around and make meaning of African languages. However, today it can be much more, and with access to even more advanced models at OpenAI, this potential is really great. Whether it is real-time translation, automated government services, voice-controlled health recommendations and teaching aids in the first language, the potential is massive.
Filling a technical gap and even making a philosophical point, too datacentric models and their local datasets are not only fair (they can even be urban-centric), but also make Orange a hell of a lot cheaper than its rivals.
