European champion no more: Mistral AI's painful bluff also eye-opener

From now on, when a European tech startup says it will surely become a valid alternative to American giants if it’s not weighed down by regulations – believe them at your own peril.

Just a few months ago, Mistral AI, a French open-source AI developer, was lobbying lawmakers in Brussels working on the now-famous AI Act to not overburden promising European AI startups with regulation.

The argument went like this: European firms need to be a competitive alternative to merciless American AI giants, and strict regulation of foundation models (large machine-learning models) would hurt their chances of success.

During negotiations, Mistral AI and Aleph Alpha, a German AI startup, were mentioned more often than not by officials in Paris and Berlin – the two capitals working on more European “strategic autonomy” and less dependence on Washington.

Competition? What competition?

The French government really pushed hard for concessions, saying Mistral AI was exceptional because it boasted an open-source model and was already valued at around $2 billion.

The talks were tough for a long time. But then, suddenly, almost unexpectedly, France, Germany, and Italy relented, and European lawmakers smoothly reached a political deal to move the AI Act forwards in the beginning of February.

Not only that. Firms working on foundational models – again, such as Mistral AI – will have to be extra transparent and share technical documents and summaries of the contents of their datasets. Almost too good to be true, right?

That’s because it might all have been a bluff. After a whole lot of posturing, and less than a month after the AI Act was rubber-stamped by the ambassadors of the EU member states, Microsoft – yes, the American tech titan – announced an investment and partnership with Mistral AI.

Monday's press release says that Mistral will make its language models available on Microsoft’s Azure platform, and in exchange, Microsoft is getting a minor stake in the firm.

Quite obviously, Mistral’s Large, a new language model supposed to directly compete with OpenAI’s GPT-4 that was also launched on Monday, will now not be open-source.

Surely, competing with OpenAI is nonsensical if both Mistral and OpenAI are in a partnership with Microsoft. It's a bit like two more or less similar products, manufactured with the help of Apple, competing in Apple's App Store.

Or like cells racing in a petri dish, the owner of which only wants her experiment to be successful – not another scientist's.

Now, the deal itself is not a finished article. European antitrust authorities will surely look into this new partnership and later decide if it doesn’t amount to a concentration of power by Microsoft.

But the smell is foul. No wonder a lot of people are now balking at the deal. The announcement in essence means that quite a few European players have been duped.

Time to rethink Eurocentrism?

Who is to blame? First, it's the ambitious government of France that has proudly but probably blindly championed Mistral AI's cause throughout the whole AI Act ordeal.

Did Paris know the startup was in negotiations with Microsoft? French politicians now say they had no clue – well, too bad because that's just poor due diligence for a wannabe European forever-leader, always eager to oppose the American kind of capitalism.

Second, it's, of course, the European lawmakers in the European Parliament, the European Commission, and the national capitals. Surely, this type of bluffing exercise must sting, and badly.

Unsurprisingly, the MEPs are fuming because Mistral AI did receive some concessions in the current version of the AI Act precisely because they argued they would be forced to cooperate with Silicon Valley giants if their wishes weren't fulfilled. Basically, the company got cake and ate it, too.

"We are extremely furious because the French government for months was making this argument of European leadership, meaning that those companies should be able to scale up without help from Chinese or US companies,” Kai Zenner, Head of Office and Digital Policy Adviser for Axel Voss, an MEP from the European People’s Party, told Euronews.

Even if we cried foul all day long, this has also been a grim reality check for Europe – the lesson is that Big Tech is big for a reason.

"They were always blaming the Parliament that we are making it kind of impossible, for those national champions, unicorns to try to compete with their global competitors."

All true. However, even if we cried foul all day long, this has also been a grim reality check for Europe – the lesson is that Big Tech is big for a reason.

Sure, Microsoft’s investment in Mistral AI is actually only $15 million – this is a tiny amount of money – and “Europe’s best hope in AI,” at least in theory, should remain independent (by the way, will Microsoft cover whatever fines to Mistral AI by European regulators in the future?).

But if $15 million is all it takes for a European tech startup to shake the hands of Big Tech moguls, maybe that’s because Big Tech is really in absolute control of the global market.

It might be too late for Europe to take on Big Tech – as long as the US federal government doesn’t take action first. The Silicon Valley firms have hoarded a huge deal of cloud computing power, so smaller startups, wherever they’re from, really have no choice but to sign deals with the giants and then try to have a piece of the pie.

Europe’s AI industry needs to either find European investors, or to look for them outside the bloc. But for that, the decision makers in Brussels might have to call for another meeting at the regulatory drawing board.