Microsoft looks to boost AI performance in European languages

The Hindu

21-07-2025

U.S. tech behemoth Microsoft is investing millions of dollars to funnel more European-language data into AI development, company president Brad Smith told AFP Monday.
With today's leading AI models mostly trained on material in English, "the survival of these languages and the health of these cultures is quite literally at stake" without a course correction, Smith said in an interview.
AI models are "less capable when it is in a language that has insufficient data," he added, which could push more users to switch to English even when it is not their native language.
Microsoft will from September set up research units in the eastern French city Strasbourg to "help expand the availability of multilingual data for AI development" in at least 10 of the European Union's 24 languages, including Estonian and Greek.
The work will include digitising books and recording hundreds of hours of audio.
"This isn't about creating data for Microsoft to own. It is about creating data for the public to be able to use," Smith said, adding that the information would be shared on an open-source basis.
The U.S.-based company has in recent months striven to position itself as especially compatible with a gathering political push for European technological sovereignty.
Leaders in the bloc have grown increasingly nervous at their dependency on U.S. tech firms and infrastructure since Donald Trump's reelection to the White House.
In June, Microsoft said it was stepping up cooperation with European governments on cybersecurity and announced new "data sovereignty" measures for its data centres on the continent.
Smith said that Monday's announcement was just the latest evidence of the company's commitment to Europe.
Most leading AI firms are American or Chinese, although Europe has some standouts like France's Mistral or Franco-American platform Hugging Face.
Away from Microsoft, some European initiatives such as TildeLM are pushing to develop local-language AI models.
The Windows and Office developer also said Monday that it was working on a digital recreation of Paris' Notre-Dame cathedral that it plans to gift to the French state, as well as digitising items from the country's BNF national library and Decorative Arts Museum.