If your eyes were turned stateside all day yesterday, you might have missed an earthquake in the artificial intelligence industry. Microsoft announced its investment in and partnership with Mistral AI, the French open-source AI developer once touted as Europe’s great hope in the global AI race. Under the terms of the partnership, Mistral will make its language models available on Microsoft’s Azure platform — including its new GPT-4 challenger Mistral Large — in exchange for Microsoft taking a minor stake in the company. What sounds like a quotidian business transaction, however, has implications far beyond either company’s bottom line. By exploring commercial collaboration with Microsoft, Mistral is endangering (if not outright trashing, as some critics say) its status as a champion of open-source AI. Mistral and the French government pushed hard for concessions for open-source foundation models in the European Union’s forthcoming AI Act, to protect European competitiveness and create a meaningful alternative to American AI giants. Now Mistral is teaming up with the biggest American titan of all, leaving the continent to question its approach to the AI race as its biggest magnet for investment flirts with the “if you can’t beat ‘em, join ‘em” approach. “In the context of [French tech summit] VivaTech coming up, and just before the EU elections in June and the fallout from them, this really does raise questions about how the European approach to tech policy might flip in the future,” said Mark Smitham, an EU policy lead at the tech policy firm Access Partnership. Europe’s competition watchdogs are already evaluating the deal between the two companies, POLITICO’s Gian Volpicelli reports. That’s even more scrutiny on Microsoft amid investigations in Europe, the United Kingdom, and the United States for its massive investment in OpenAI and brief hiring of its CEO Sam Altman. For some critics, the deal between Microsoft and Mistral is less a surprise than an outright betrayal. Kris Shrishak, a fellow at the Irish Council For Civil Liberties and a close observer of the AI industry, told me last December that Mistral had a chance to provide a real, meaningful alternative to commercialized American AI dominance — “If they hold up their end of the bargain of offering good, free, open-source models.” These days, he’s not sounding so sanguine: “Mistral's bluff is now out in the open,” he wrote DFD in an email in which he accused Microsoft of attempting to skirt competition regulation. (Microsoft and Mistral did not immediately respond to a request for comment.) “Microsoft is aiming to extinguish competition in the AI ecosystem,” he added. “The only AI competition remaining will be within Microsoft's Azure marketplace. We have seen the pitfalls of this previously with Apple’s App Store. They don't serve the public interest.” One huge concern for European critics of the deal is whether it was negotiated while Mistral was actively lobbying for carve-outs in the forthcoming AI Act. Kai Zenner, digital policy adviser for European Parliament member Axel Voss of Germany, told the European news outlet EuroNews Next that there “are certain red lines” around such lobbying, and that he suspects the French government was as surprised by the deal as many in the tech community. (A spokesperson from France's economy, finance and digital ministry told EuroNews Mistral is “a source of pride for France and Europe.”) “The narrative for a long time in Brussels has been around the concept of European sovereignty,” Smitham said. “This really calls into question that fundamental approach.” That leaves European regulators uneasy vis-a-vis their traditional role as a regulation-forward counterweight to America: If the Eurocentric approach that guided the final AI Act negotiations in Brussels ended with Mistral teaming up with an American colossus, it might be time to go back to the regulatory drawing board.
|