DE — heise online

Mistral and DeepL Chiefs Warn Against Europe's AI Surrender

AIOctober 25, 20262 min read

Top European tech bosses, including the heads of Mistral and DeepL, are warning against a debate over whether Europe should build its own advanced AI models. They say giving up on this would leave the continent dependent on others.

A fight over Europe's AI strategy

Some European leaders, including EU Commission President Ursula von der Leyen, think Europe does not have to lead in the most advanced AI models. They say Europe can still earn money from AI by building useful apps. Other tech leaders strongly disagree. They warn that this path would be a kind of surrender.

Who is speaking out

Arthur Mensch leads Mistral, a French AI company. Jarek Kutylowski leads DeepL, a German AI translation company. Mensch wrote a LinkedIn post called “Build or surrender”. He said that if Europe stops building its own frontier models, it will lose its importance in the world of AI. Max Junestrand, head of the Swedish legal-tech company Legora, sparked the debate by saying that building a competitive frontier lab looks like wishful thinking right now.

Key facts

The European Commission started building seven AI gigafactories, with a total value of about 30 billion euros.
An AI model is a computer program trained on huge amounts of data. Training and running the most advanced ones needs a lot of computing power and money.
A group of European companies led by the Italian firm Domyn is working on its own frontier model. It should be available in all official EU languages.

Why it matters

The bosses fear that Europe could depend on AI tools from other countries. They also worry about a trade deficit, which means Europe would spend more money on foreign technology than it earns from it. Kutylowski says that many software companies now spend a large part of their budgets on AI models instead of engineers' salaries. Without European models, that money would flow to the United States.

What to watch

Whether the seven AI gigafactories are built on time and with the planned money.
How far the Domyn-led group gets with its frontier model.
Whether EU leaders change their plan to focus on AI apps instead of top AI models.

Glossary

Mistral: A French AI company that builds large AI models.
DeepL: A German company that makes AI-based translation tools.
Legora: A Swedish legal-tech company that makes software for lawyers.
AI gigafactories: Very large computing centers for building and running AI, planned by the EU.
Domyn: An Italian company that leads a European group working on a frontier AI model.

Same story, different eyes

The StrategistAI

Europe's AI debate is really about who sits at the top of the supply chain

Seen through a tech-geopolitics lens, the 'build or surrender' argument is about where dependency sits in the stack. Whoever owns the frontier models sets the terms, pricing and access for everyone building applications on top, so a purely application-focused strategy can feel efficient yet leave Europe exposed to decisions made elsewhere. The other side has a fair point too: compute is costly and capital is limited, so the seven AI gigafactories (about 30 billion euros) and the Domyn-led multilingual model look like Europe trying to hedge between both paths. How the gigafactories' compute will actually be shared among European model builders is not yet known.

The SkepticAI

Build or buy: which numbers would settle Europe's AI debate?

Both sides make a reasonable case, and I think the debate gets clearer if we ask what we could check. Kutylowski's example is that more software budgets now go to AI models instead of engineer salaries. If we could see how much of that spending actually flows to non-European providers, the trade-deficit worry would be much easier to weigh. On the other side, it would help to know what a competitive frontier lab costs compared with the roughly 30 billion euros planned for the AI gigafactories, and how the Domyn-led consortium's model performs once it is released. Neither is known yet.

The FounderAI

Europe's 'build vs. apply' split opens a gap for sovereign-model builders

Seen through a founder's lens, this debate leaves room on both sides. If Europe leans on foreign models, a gap opens for startups that make those apps trustworthy, multilingual and compliant, and the gap that closes is any moat built purely on renting someone else's model. The fact that a big share of software budgets is moving from engineers' salaries to model spend, as DeepL's CEO points out, makes the case for starting now: a first step could be an app that can swap between models, including the EU multilingual model Domyn's consortium is building, so you aren't locked into one supplier. How much of the 30 billion euro AI gigafactory money will reach small teams is not yet known.

OpenAI Holds First Korean 'Model Masters' Event to Support Startup AI Development

OpenAI held its first 'OpenAI Model Masters' event in Korea on the 23rd at its Seoul office, where 11 startups demonstrated services built on its models. The program, which provides API credits to startups recommended by venture capitalists, aims to help turn AI ideas into real products. OpenAI highlighted the selected startup Cutback's video-editing work as a standout example.

AIITmediaOct 26

Block, led by Jack Dorsey, Unveils Dessa, a Full-Duplex Voice Model Trained Only on Speech: "Feels Like Talking to a Person"

Block (formerly Square) announced Dessa, a full-duplex speech-to-speech model trained solely on human conversational audio without any text, as a research preview on October 23 (local time). Jack Dorsey said it is the most realistic full-duplex voice model the company has built, and Block says it runs locally on a MacBook Pro with M4 Max or higher.

AITechCrunchOct 26

Grok releases the first seven minutes of its AI-generated ‘Odyssey’

Three months after Elon Musk promised a full-length AI-generated version of “The Odyssey,” Grok has released seven minutes it calls “Episode 1 of our Odyssey series.” The footage was made with Wonder Studios and director Dev Walker, and the reviewer found it less ugly than expected but lacking in artistry, momentum, and human interest.