How is Orange Using OpenAI Tech For Translation Services?

Share this article
Share this article
Prioritise Us on Google
Credit: Orange
Telecommunications giant Orange is set to use OpenAI’s latest AI models and fine-tune language models to translate regional African languages

French mobile operator Orange has begun working with OpenAI’s pre-release AI models and fine-tuning large language models to translate regional African languages.

Building on a deal signed with OpenAI in 2024, the new models can extend Orange’s current work to far more complex uses.

Orange and OpenAI Partnership (Credit: Orange)

Orange’s operations in Africa

While headquartered in France, Orange operates extensively around the world, including serving more than 140 million customers across 18 countries in Africa. 

Its strategic focus across the continent is to expand digital inclusion and enhance connectivity by investing in mobile networks, 4G expansion, fibre optics and digital services to meet growing demand for internet access and financial inclusion.

The company is particularly active in mobile money through Orange Money, a widely used service across Africa, enabling millions to transfer funds, pay bills and access essential financial services – even in remote areas. 

Orange also supports innovation hubs and digital skills training through initiatives like Orange Digital Centres.

Youtube Placeholder

How is Orange working with OpenAI in Africa? 

With the benefits of AI models largely bypassing African languages, mainly due to challenges such as a lack of data and limited computational resources, Orange is set to change this alongside OpenAI. 

With access to OpenAI’s pre-release AI models and fine-tune large language models, Orange will utilise these AI tools alongside its collected samples of African regional languages to translate and deploy locally using OpenAI’s open-weight reasoning models – 'gpt-oss-120b' and 'gpt-oss-20b'.

“This collaboration with OpenAI is foundational to our strategy of using state-of-the-art AI models that are both trusted and responsible,” says Steve Jarrett, Chief AI Officer at Orange. 

Steve adds: “This strategy drives new use cases to address sensitive enterprise needs, help manage our networks, enable innovative customer care solutions including African regional languages, and much more. The most advanced AI, like these new OpenAI open models, and our world-class team of AI experts, allow us to balance value generation while mitigating cost and environmental impact, in line with our Responsible AI approach.” 

Steve Jarrett, Chief AI Officer, Orange

What are open-weight models?

OpenAI’s open-weight models are machine learning models with pre-trained weights – the values learned during training – that are publicly available for anyone to download, use or fine-tune. 

While they are not fully open-source, these models offer transparency and flexibility to users, enabling researchers and developers to run them locally, adapt them for specific tasks or integrate them into products without relying on OpenAI’s API.

Examples include Whisper (speech recognition), CLIP (image and text understanding) and GPT-2, allowing experimentation and innovation across industries.

Brad Lightcap, Chief Operating Officer at OpenAI, adds: “Our ongoing work with Orange shows how forward-thinking businesses can use our open models to solve real-world problems—from boosting network efficiency to improving African language support, they are delivering meaningful benefits directly to their customers.”

Brad Lightcap, Chief Operating Officer at OpenAI

The digital divide and language barriers in Africa

Africa’s digital transformation has made significant strides in recent years. However, much of the continent remains underserved by AI technologies. 

A key barrier is language. With more than 2,000 languages spoken across Africa, many of which are not represented in digital datasets, mainstream AI models have largely bypassed the continent. 

These models are typically trained on dominant global languages such as English, Mandarin and Spanish, leaving vast populations excluded from the benefits of voice recognition, translation and generative AI tools.

Excluding these languages increases the digital divide, limiting access to vital services. 

Without support for local languages, millions are unable to fully engage with digital platforms, from healthcare and education to financial services. Orange’s collaboration with OpenAI seeks to change this by laying the groundwork for more inclusive AI solutions.

By addressing the data and resource gaps, Orange and OpenAI are enabling communities to access information, services and opportunities in the languages they understand best – a vital step in closing the digital divide across the continent.

Company portals