The EU AI Act Is Here. We’re Screwed

I’m sure you’ve been eagerly awaiting another rant, especially about this sort of stuff. But before I start ranting, let’s actually start with some facts! Here’s what Thierry Breton of the European Commission thinks about the AI Act:

Okay. So he thinks regulating AI creates “a launchpad for EU startups”.

I’ll just leave that there for a moment – similar how you just nod and smile when a completely confused person approaches you on the street and starts talking nonsense.

So maybe he’s just confused. Let’s move on and see what other people have to say. People who were actually involved in the EU AI Act.

Here’s the co-founder of Mistral, Europe’s pretty-much-only serious AI company:

[…] Recent versions of the AI Act started to address ill-defined “systemic risks”. […] The AI Act comes up with the worst taxonomy possible to address systemic risks. […] The intention of introducing a two-level regulation is virtuous. Its effect is catastrophic.

Arthur Mensch, co-founder of Mistral AI, on X / Twitter

So the TLDR is that he initially thought it could be good, but then it evolved into a complex monster and its current incarnation is terrible.

But maybe, while the EU commission dude might be confused, Arthur Mensch might be biased. After all, he’s working at an AI company (the only one in Europe) which might present a conflict of interest.

So let’s move on and see what the lead author of the AI Act, Gabriele Mazzini, has to say:

The regulatory bar maybe has been set too high. […] There may be companies in Europe that could just say there isn’t enough legal certainty in the AI Act to proceed.

Gabriele Mazzini, lead author of the EU AI Act. Source: Bloomberg interview

So even the lead author of the EU AI Act isn’t very happy with it.

But maybe, while the EU commission dude might be confused, Arthur Mensch of Mistral might be biased, this dude might be.. umm.. maybe Bloomberg kidnapped his cat to extract that quote from him.

Yeah, maybe those people are wrong and the EU AI Act is great. Right? Right?

Let’s see what’s happening in the real world. You know, real companies building innovative AI products. In other words, companies outside of the EU.

A while ago, Apple announced some new LLM / AI features coming to iOS. They’ll be called “Apple Intelligence” and Apple won’t ship them to customers in the EU.

Cool.

So if the EU AI Act was supposed to be a “launchpad for EU startups”, I guess it might have malfunctioned for Apple, because Apple’s launch was cancelled and they decided to launch in the US instead.

But maybe Mistral, Europe’s only hope, will be able to use this magic launchpad?

Let’s just say that things might get complicated for them. France is considering fining Nvidia over anti-competitive practices regarding their GPUs and CUDA, their programming language. The fine would be up to 10% of Nvidia’s global turnover – if this happens, it would be much more rational for Nvidia to exit the French market than to comply with it.

Now, imagine that – Mistral, Europe’s only AI company, would no longer have access to Nvidia GPUs.

Historic!

So even if we have this magic launchpad for EU startups, we might have neither the rockets, nor the fuel. Only a piece of concrete. I wonder what you’d do with that. Maybe all the new EU AI Act auditors can organize a barbecue there, because they won’t have much to audit anyway. While they’re there, they could also search for Gabriele Mazzini’s kidnapped cat.

---

On a (just slightly) more serious note, I see two scenarios for companies:

Companies which are currently on the EU market will very carefully assess whether they ship new features to the EU. They might follow Apple and ship a “dumbed down” version of their software.

Companies which are currently not on the EU market (this includes startups) will very carefully assess whether to enter the EU market at all. The AI Act is just another bureaucracy headache on top of GDPR and other regulations which keep founders from building their products. Incorporating outside of the EU, e.g. in Singapore (like Ghost did), might become more popular.

Zooming out, the bigger picture here is that the EU is turning into a micro-manager by regulating an increasingly large number of things, in an increasingly unreasonable manner:

• The Nvidia / CUDA case in France as mentioned above – this is essentially telling Nvidia “hey, you guys did some crazy good innovation with the CUDA programming language, but you should open it up for everybody and make no money off it!”.
• Meta’s “Pay or Consent” case where the EU says “letting users choose between ads or a subscription is not good. make it free for everyone, without ads!”, essentially telling Meta to not make any money at all.
• The Apple App Store case where the EU is telling Apple “you shouldn’t be running a closed-off app ecosystem and make tons of money off it. instead, open it up for everyone and make no money!” (see a common theme here?).

I’ve certainly over-simplified, and each case has its own set of details which may or may not make regulatory intervention seem reasonable.

What’s clear, however, is that it’s becoming increasingly hard for tech companies to do business in the EU at all. And the EU isn’t a big market for those companies – for Meta, only 10 percent of its revenue comes from the EU, and for Apple, it’s only 7 percent (link to awesome Stratechery post). The EU-imposed fines, however, are based on global revenue, so they’re much larger. That’s crazy!

It might become a real thing that tech companies exit the EU market. And in contrast to China, we don’t have a homegrown tech industry to step up.

That, indeed, would be.. historic!