French telecom giant Orange is turning its focus to Africa’s linguistic diversity, announcing plans to fine-tune OpenAI’s latest AI models to work with the continent’s wide range of regional languages.
Why it matters
Over 2,000 African languages have historically been left behind in the development of AI systems due to data scarcity and limited computational resources. Orange’s initiative could mark a major step toward changing that.
Driving the news
- Orange, which operates in 18 African countries, signed an agreement with OpenAI last year, granting it early access to experimental AI models.
- The company began using OpenAI’s Whisper speech model earlier this year for African languages, but is now expanding to more complex large language model (LLM) tasks.
- Using OpenAI’s new open-weight models, Orange can fine-tune the tools with its own African language datasets—without needing to start from scratch.
What they’re saying
“We plan to provide the fine-tuned models for free to local governments and public authorities,” said Steve Jarrett, Chief AI Officer at Orange.
“This initiative is a blueprint for how AI can help bridge the digital divide,” he added, emphasizing the role of local partnerships in building inclusive AI systems.
What’s next
- Orange aims to deploy the customized models locally and make them available to startups, communities, and governments.
- By supporting translation, communication, and access to services in native languages, the move could help bring millions of Africans online and into the digital economy.
The big picture
A Cornell University study and a report by Nature have highlighted how African languages have largely been overlooked in mainstream AI development.
Orange’s project could serve as a model for how big tech can partner with local actors to make AI work for underrepresented regions.
Source: Reuters
