| One of the biggest divides in the AI world is getting a high-profile test case in Europe. As POLITICO’s Mohar Chatterjee and Gian Volpicelli report, France is putting its chips down on open-source AI, hoping to position itself as a global leader in the AI industry. The moves France is making are significant because they intersect with two of the biggest global stories in AI right now: Europe’s role as its chief regulator, and the growing philosophical divide between open-source developers and closed shops like the (now-ironically-named) OpenAI. Believers in open-source AI development — who think AI tools should be totally transparent and accessible to the public, although “the public” might occasionally be narrowly defined — see it as a way to democratize the most powerful new technology to come along since the internet itself. Companies like Google and OpenAI see that power as precisely why AI tools should be closely held until they can be thoroughly safeguarded against potential misuse, however they choose to define it. As for the French, they see siding with the former as a way to kill two, or even three, birds with one stone. By championing open-source AI systems, they can fight U.S.-based tech giants, bolster their own tech sector and maybe even carve out a new regulatory lane as the European Union’s AI Act hurtles toward passage. It’s a bold, possibly far-fetched experiment, but if it pans out it could lead to a far more multipolar new era of tech development than the current one, dominated by just a handful of U.S. titans. Europe is largely backing the push, as Kai Zenner, a European Parliament policy adviser, told Mohar and Gian: "We completely agree with the French assumption: We see open-source AI as a big chance… If Europe really wants to catch up with the United States and China in AI, then without drawing on models or data sets from the open-source community, we would never have a chance.” In June France’s President Emmanuel Macron announced the country would invest €40 million in French-made generative AI projects, and €500 million for a broader slate of AI projects Macron said was aimed at creating market and research “champions” in France. Mistral.ai, a Paris-based startup, raised €105 million this summer for its attempts to build an open-source competitor to ChatGPT, as state-run efforts face the typical roadblocks to bloc-wide coordination. Coincidentally, France’s effort to shift the playing field away from stateside AI giants has a lot in common with what one of the latter is attempting. Meta made a huge splash in February by announcing that the company believed open-source AI was the future, with its chief AI scientist Yann LeCun — a Frenchman — telling the New York Times “The platform that wins will be the open one.” French AI talent runs deep, Mohar and Gian point out: the country has a robust academic network and some of the EU’s highest-quality AI labs. But… it’s still not America, and the French startup ecosystem faces significant disadvantages in academia and cloud computing power. That’s not the only thing standing in France’s way. For one, the actual open-source industry is still unclear on what exactly the regulatory environment might look like once the AI Act takes effect, with Github urging more clarity from the EU just a few weeks ago. They also share the concern with companies like OpenAI that “foundation models” underpinning popular AI software tools could face separate regulatory obligations — something Nick Clegg, Meta’s president of global affairs, told Mohar and Gian “seem[s] to impose potentially some pretty complex and potentially somewhat unworkable conditions on open-sourcing large language models altogether.” More than anything else, France’s ambitious push might reflect exactly how much still seems up for grabs in this feverish moment of AI development, as some of the planet’s biggest players both private and public attempt to zig where other companies are zagging. “Europe has to do whatever it can to be part of the game,” Cedric O, a former French digital minister who is now a shareholder and adviser to Mistral.ai, told Mohar and Gian — and by doing that it’s helping to “define” along the way exactly what that game is, depending on which model of AI development ultimately wins out.
| 
