Microsoft to enhance AI performance for European languages

• Microsoft invests millions to boost AI with European-language data, preserving cultural diversity.
• Strasbourg research units to digitize books, record audio for 10 EU languages.
• Open-source data initiative supports public use, aligns with European technological sovereignty.

Microsoft is investing millions to enhance AI development with European-language data, says company president Brad Smith in an interview on Monday. The initiative aims to preserve linguistic diversity and cultural health in Europe, where leading AI models, primarily trained on English data, underperform in other languages.

“The survival of these languages and the health of these cultures is quite literally at stake,” Smith emphasized, noting that AI models are “less capable when it is in a language that has insufficient data.” This deficiency could drive users toward English, sidelining native languages.

Beginning in September, Microsoft will begin the process of building research centers in Strasbourg, France, to increase multilingual data within AI to at least 10 of the EU so that it is available in 24 of EU languages, such as Estonian and Greek. The project entails the digitization of books and hundreds of hours of audio recording, and all the information will be presented on an open-source basis.

“This isn’t about creating data for Microsoft to own. It is about creating data for the public to be able to use,” Smith stated.

Also Read: China Introduces First Humanoid Robot with Self-Swapping Battery

This move aligns with Microsoft’s broader commitment to European technological sovereignty, amid growing concerns over reliance on U.S. tech firms, especially following Donald Trump’s reelection. In June, Microsoft announced enhanced cybersecurity cooperation with European governments and new data sovereignty measures for its European data centers.

Smith highlighted that this initiative is further proof of the company’s dedication to the region.

Beyond AI, Microsoft is working on a digital recreation of Paris’ Notre-Dame cathedral, to be gifted to the French state, alongside digitizing collections from France’s BNF national library and Decorative Arts Museum. Meanwhile, European AI initiatives like TildeLM and companies such as France’s Mistral and Franco-American Hugging Face are also advancing local-language AI models.