Telecoms operator Orange has announced further collaboration with ChatGPT creator OpenAI that includes incorporating several African languages into artificial intelligence (AI) models.

This is part of a bigger announcement from Orange that it will deploy OpenAI’s new advanced open-weight reasoning models, “gpt-oss-120b” and “gpt-oss-20b,” in its infrastructure.

It said this will make customers’ data even safer and meet the growing demand for sovereign, state-of-the-art AI solutions.

“As an early access partner, Orange is one of the first companies globally to deploy OpenAI’s new open models,” the French operator said in a statement.

Inclusive AI for Africa

Orange said a core focus of the collaboration was driving digital inclusion and AI innovation across Africa and accelerating the use of African languages in AI models.

The underrepresentation of African languages in large language models (LLMs) remains a barrier to AI adoption on the continent, with only a small fraction of Africa’s over 2,000 languages being supported.   

OpenAI’s gpt-oss models will enable Orange’s customers across its 18-country footprint in Africa and the Middle East to communicate naturally in their local languages with customer support, sales and marketing.

“In the future, Orange will release the customized AI models in open source for free, to local government authorities to use across its public services,” the telco said.

Orange’s operations in Africa and the Middle East are a key priority area for the French telco which last month said the region was the main contributor to revenue growth over the first six months of its financial year.

OpenAI’s new models will enable Orange customers in Africa to communicate naturally in their local languages with customer support, sales and marketing. (Source: Freepik – AI-generated)

Steve Jarrett, chief AI officer at Orange, said the collaboration with OpenAI will drive new use cases to address sensitive enterprise needs, help manage its networks and enable new customer care solutions, including African regional languages.

Brad Lightcap, chief operating officer at OpenAI, said the tie-up shows how businesses can use open models “to solve real-world problems” like boosting network efficiency and improving African language support.

‘Frugal’ AI approach

Orange said it had a “frugal” AI approach – aiming for high performance with less computing power, less data and less energy as part of what it calls its “Responsible AI framework.”

Jarrett explained that the telco was looking to balance value generation while mitigating cost and environmental impact.

Orange said it can run the OpenAI models in its own data centers and directly manage and optimize the power consumption of the customized models to reduce emissions.

“By maintaining control over the deployment environment, Orange can host AI workloads locally across any of its 26-country footprint [globally], whilst safeguarding sensitive data and complying with diverse and evolving national regulations across Europe, Middle East, and Africa,” it said.  

New applications and services

Orange said the deployment of OpenAI’s open-weight reasoning models will also help it develop new applications and services for customers.

This includes natural language understanding in multiple languages for voice assistants and chatbots.

It will use AI to improve the quality and resilience of its networks and achieve higher levels of network automation.

It also sees an opportunity to offer custom AI solutions to its enterprise customers such as AI chatbots, advanced voice recognition-based services and AI analysis of sensitive data. 

Orange has already been using AI to optimize its networks in several African countries.

Former Orange Middle East and Africa deputy CEO Brelotte Ba told Connecting Africa at the end of 2024 that the company was investing in AI to increase automation and improve customer service.

Orange’s AI-based “Smart Capex” solution was first piloted in Spain and has now been rolled out in Côte d’Ivoire, Senegal, Mali and Jordan to help optimize network capacity and streamline investment.