For years, the European Union has been quicker to regulate artificial intelligence than invest in it. Now, with the battle for AI supremacy heating up thanks to a powerful new open source model from China’s DeepSeek built for a fraction of the price of its U.S. rivals, it’s stepping off the sidelines. This morning Brussels announced plans to develop an open source AI model of its own, with $56 million in funding to do it.
The investment will fund top researchers from a handful of companies and universities across EU countries as they develop a large language model that can work with the trading bloc’s 30 languages. The project will also tap into supercomputers from companies like Spain’s Mare Nostrum and Italy’s Leonardo, both of which have received funding from the EU.
The budget is a tiny fraction of the money pouring into leading American AI labs like OpenAI, which is reportedly raising up to $25 billion at a $300 billion valuation, or even one of Europe’s hottest AI unicorns, Mistral, which raised $640 million last year. But given DeepSeek’s breakthrough, what once sounded like chump change in the AI space could go much, much further: the EU’s backing for the project is 10 times what Deepseek claims to spend on training its own R1 model. “I think Europe should be now seen to be in the race again,” says project coordinator Jan Hajič, professor of computational linguistics at Prague’s Charles University.
The project aims to fund an open source LLM that European companies and governments can build on top of, assured that it was built with the EU’s values baked in. Open source models can allow users to make tweaks and modifications while OpenAI and Anthropic’s AI are walled gardens. “There’s a need for open source models that are aligned in terms of languages, values and with society at large,” says Peter Sarlin, cofounder of Finnish AI lab AMD Silo. (Mistral, which is based in Paris and was valued at over $6.2 billion last year, has made a similar pitch, but it’s not participating in the project.)
Europe has good reason to build its own LLM: Meta, OpenAI and other U.S.-based companies have slowed or blocked the release of tools in Europe because of legal risks stemming from its AI Act, which seeks to ban “high-risk” uses of AI and takes effect in February. Meta founder Mark Zuckerberg and Spotify cofounder Daniel Ek warned that “incoherent and complex regulation” could lead to Europe falling further behind on AI in an open letter published last year.
Since first passing the law last summer , the EU has made some moves to bolster its tech sector. Its annual budget for research grants was bumped a quarter to $1.5 billion. It will also invest $1.5 billion into upgrading the continent’s network of supercomputers, and last week EU Commission president Ursula von der Leyen unveiled plans meant to stoke growth among its 27 member nations.
For some, the project is reminiscent of a failed $550 million project to build an European search engine to rival Google. Why invest public money, when venture capitalists have been vying to back private projects? “Sprinkling money around in a fragmented fashion is more about PR than driving actual results and impact,” warned tech investor Rasmus Rothe of Merantix Capital.
More from Fo
Read the full article here
