Today, we publish a Transformative AI Strategy for Europe.
Over the last few months, we’ve rallied researchers and engaged with governments to develop a plan for protecting the prosperity, sovereignty, and security of Europeans in a time of rapid AI progress.
https://t.co/Jc34ptCRMc
@MonikaSchnitzer and I are honored to have convened an all-star team of thinkers and researchers contributing ambitious near-term objectives to make Europe relevant again, including:
— Creating a Member State Alliance for Supply Chain Security (@anton_d_leicht et al)
— Making European institutions ready to act in a transformative AI world (Conor McGlynn et al)
— Securing Europe's share of global AI compute, in a ‘European Way’ that benefits local communities (@philip_fox_ et al)
— Ensuring resilience to AI crises (@ben_s_bucknall et al)
— Making Europe the global leader in assurance technology
— And more objectives around security of supply and leverage (@milorignell et al), economic strength (@FraukeStehr et al), and safety/security (@NoemiDreksler et al)
Our all-star senior expert council of Europe’s best and brightest (and non-European friends) reviewed drafts, provided strategic advice, and made suggestions for how to make the strategy more useful: @Ph_Aghion@Christophkw@bakkermichiel@ischinger@vestager@aleks_madry@FuestClemens Marta Kwiatkowska @LeoVaradkar@antonosika@DAcemogluMIT@Yoshua_Bengio
It’s never been more clear: AI is real, and Europe needs to act. We’ve had all the warning shots and wake-up calls we need. Now the question is ‘what must be done?’ This strategy is our answer.
https://t.co/Jc34ptCRMc
Great talking about AI with @eucopresident António Costa today. And even better that he invited European leaders to discuss AI this week at an informal EUCO in Dublin. Europe needs to act fast. https://t.co/rZNntAdsZP
Excellent discussion today with the researchers behind the report “A Transformative AI Strategy for Europe”.
Europe must use AI to reinforce its competitiveness and sovereignty, while keeping the technology safe and under human control.
This week at the informal #EUCO in Dublin, Leaders will have a strategic discussion on AI.
#phoenixperönlich | „Europa muss aufwachen“ sagt der KI-Experte Daniel Privitera @privitera_ vom @kira_center_ai mit Blick auf die aktuelle Diskussion um Künstliche Intelligenz.
Jetzt streamen 📱 https://t.co/WRnSnXHnXA |📺 Morgen 11.30 Uhr bei phoenix
KI-Skepsis ist jetzt Mainstream. Doch der Weg zu effektiver Regulierung ist noch weit.
Meine Kolumne im Handelsblatt -- mit Verweis auf die diese Woche veröffentlichte "Transformative AI Strategy for Europe" von Monika Schnitzer Daniel Privitera und Kollegen.
Great to see @vonderleyen put young people at the centre of the AI and jobs debate, and note carefully that overall unemployment remains low.
The Quality Jobs Act, as announced, focuses on working conditions and algorithmic management. It doesn't improve Europe's ability to see AI's labour market impact: no vacancy tracking, no AI displacement module in the Labour Force Survey.
The speech also lacks flexicurity: letting companies restructure quickly while giving workers real support to change tracks.
Good to see more on middle-power coordination, as we suggested in the Transformative AI Strategy convened by @MonikaSchnitzer and @privitera_, where I wrote the labour market section. It has much more on verification and assurance: https://t.co/zP4tc8ccpc
How much did President @vonderleyen big speech today align with our Transformative AI Strategy?
Tldr: The President set a high bar for herself and the Commission with this speech, and I’m glad she did. Our strategy urged for far-reaching action before the end of the year, and today the President opened the door to that.
Point-by-point comparison of her speech with the most urgent objectives in our Transformative AI Strategy:
1) Creating a Member State Alliance for Supply Chain Security:
No mention of Member States (which wouldn’t be for her to announce anyway) and no explicit mention of the chip supply chain. Instead, an intention to “step up together with our like-minded partners like Canada, the United Kingdom and others” on “AI security”. This should leave the door open for both “security” in the “protecting against risks from AI” and in the supply chain security sense.
2) Making Europe’s institutions ready to act in a world with transformative AI:
Completely absent. The President wants to get AI right, with a special focus on frontier AI impacts. Frontier AI experts should therefore directly inform the political leadership on all necessary actions.
3) Securing Europe’s share of global AI compute, in a ‘European Way’ that benefits local communities:
“We need to massively boost our computing capacity. And we will set out our ideas for how to achieve this in a clean and efficient way.” Directionally good, but announcing a 15% target of global AI compute would have been better. The setting out of ideas, and the acting, needs to happen asap. We describe several easy, quick wins in IO-3 of our strategy.
4) Ensuring resilience to AI crises:
Mentions cyber and loss-of-control risk, pacing the frontier, recursive-self-improvement. That’s good and shows situational awareness! Actions announced/hinted include “We want to team up on model evaluation, verification, early warning, AI security and much more”. That’s the right direction, and it needs to happen fast.
5) Making Europe the global leader in AI assurance technology:
Explicit mention of “verification”, which is at the core of our objective in the strategy. Good!
Overall: This speech was significantly stronger on AI than previous speeches.
But it will only mean something if we will see the Commission approach these topics with the speed, seriousness, and prioritization appropriate for a “tipping point” topic, from tomorrow onwards. A lot of people will have their doubts about that, and I hope the Commission will prove them wrong. IO-2 in our strategy is full of non-costly actions that would make this possible.
Overall, the key question from tomorrow until the end of the year is: Will the Commission switch gears and deliver on today’s announcements?
Speech: https://t.co/XdlfQSOBH1
Our Transformative AI Strategy: https://t.co/I7KGCNSoXs
Today, we publish a Transformative AI Strategy for Europe.
Over the last few months, we’ve rallied researchers and engaged with governments to develop a plan for protecting the prosperity, sovereignty, and security of Europeans in a time of rapid AI progress.
https://t.co/Jc34ptCRMc
@MonikaSchnitzer and I are honored to have convened an all-star team of thinkers and researchers contributing ambitious near-term objectives to make Europe relevant again, including:
— Creating a Member State Alliance for Supply Chain Security (@anton_d_leicht et al)
— Making European institutions ready to act in a transformative AI world (Conor McGlynn et al)
— Securing Europe's share of global AI compute, in a ‘European Way’ that benefits local communities (@philip_fox_ et al)
— Ensuring resilience to AI crises (@ben_s_bucknall et al)
— Making Europe the global leader in assurance technology
— And more objectives around security of supply and leverage (@milorignell et al), economic strength (@FraukeStehr et al), and safety/security (@NoemiDreksler et al)
Our all-star senior expert council of Europe’s best and brightest (and non-European friends) reviewed drafts, provided strategic advice, and made suggestions for how to make the strategy more useful: @Ph_Aghion@Christophkw@bakkermichiel@ischinger@vestager@aleks_madry@FuestClemens Marta Kwiatkowska @LeoVaradkar@antonosika@DAcemogluMIT@Yoshua_Bengio
It’s never been more clear: AI is real, and Europe needs to act. We’ve had all the warning shots and wake-up calls we need. Now the question is ‘what must be done?’ This strategy is our answer.
https://t.co/Jc34ptCRMc
Thoughts on the economic measures and recommendations in the TAI Strategy for Europe.
It calls for institutional capital for growth companies, pension-fund investment, portable employee equity, easier restructuring and a Single European Growth Market.
None of this can work at European scale without common infrastructure underneath it.
That infrastructure is EU Inc (@euinc_petition).
We are still trying to build European capital markets on top of 27 different company systems. Capital may be European in ambition, but the legal structures, registers and rules it has to move through remain fragmented.
EU Inc is often treated as a startup-law project, or simply an investor-ready structure that makes fundraising easier. Its economic role is MUCH BIGGER.
Europe needs a common corporate vehicle that investors, lenders, employees and markets can understand and trust across borders. One that supports a company from formation through different stages and forms of capital, rather than forcing it back into national fragmentation when it expands.
This matters for all growth companies, but especially for the capital-intensive companies and projects Europe says it wants to build. Manufacturing, energy, defence, compute and other industrial sectors cannot be financed by venture capital alone. They need long-term institutional capital, credit and lenders who understand the structure and can assess the risk with confidence.
This is why the technical points matter. Be it insolvency, employee equity or the central register, these are not separate legal details. These are precisely the aspects which create confidence in the market.
Capital providers need to understand the upside, but they also need to know what happens when a company fails. Today that still means navigating different insolvency systems, processes and outcomes depending on the country. If investors and lenders cannot predict the downside, they cannot price risk confidently. And if they cannot price risk, capital does not flow, especially the more patient and varied forms of capital that industrial companies need.
The same applies across the company lifecycle. One authoritative source of company data gives banks, investors and counterparties something they can rely on. Employee equity that works across borders and remains useful as companies scale allows European businesses to compete for global talent. Cross-border mobility means a company does not have to rebuild its legal structure every time it grows into another market.
Together, these elements make the EU Inc vehicle understandable and dependable across Europe. They build lender confidence, support institutional investment, improve liquidity and allow more types of growth capital to enter the market.
Without this common infrastructure, every new capital initiative remains a separate national intervention. Capital stays fragmented. Liquidity stays thin. Companies still face 27 systems. Europe cannot build functioning capital markets at scale that way, and it cannot industrialise on venture funding alone.
Mario Draghi called for a new EU-wide legal statute for innovative companies. Enrico Letta proposed a simplified European company. The TAI Strategy now sets out AGAIN the economic measures Europe needs to build and scale its companies.
EU Inc is the CONCRETE vehicle that connects those recommendations and makes them operational.
This is why the current negotiations matter far beyond startup policy. If EU Inc is reduced to an empty label on top of national fragmentation, it will not do the job Europe needs it to do.
We need to maintain the ambition and withstand the pressure to pull it back towards 27 national systems.
Reindustrialising Europe is not only about financing factories, energy, defence or compute but it also requires the institutional infrastructure behind them: a legally certain, predictable pan-European corporate vehicle through which capital can be deployed, risk can be priced and companies can scale. EU Inc is part of that market infrastructure. We need to build it properly but it's currently under pressure.
I agree it is encouraging to see the EU taking the stakes of advanced AI seriously, I applaud VdL and her advisors for understanding this.
The key is to now turn these statements into urgent and meaningful action. The recent Transformative AI Strategy for Europe identifies the right priorities and next steps: https://t.co/HBUheHnEAD
On 'pacing' specifically, the first steps are domestic, inside the frontier labs themselves (e.g. embedded auditors). But in the (possibly near) future, frontier companies and their domestic governments could want optionality to pace internationally. The technical tooling to support this does not yet exist. This is where the EU can lead.
One concrete idea:
Turn an existing or forthcoming AI factory into the world’s first AI verification accelerator.
A standalone programme covering R&D, datacentre retrofits, and connecting builders with nation-state-grade red-teaming and operations.
Steps:
- Task the AI Office to hire a small team (a TPM, an ML engineer, a hardware-security engineer and an
infrastructure engineer) to run it.
- Run a retrofit programme to equip it with security and networking components for verification.
- Run access programmes so builders can experiment on the cluster.
- Put IP-sharing arrangements in place with startups that use it.
- Give it priority access to government red-teaming (e.g. ENISA).
Indicative cost: ~€25–35M over two years for a ~1 MW cluster (~600–800 GPUs)
As I work for a startup building AI verification tech, we understand that there is a critical role for governments to reduce the time between technical designs through to adversarial red teaming with nation-state capabilities. An AI verification accelerator in Europe would be a great way to do this
I recently spent some time thinking about what Europe should do regarding AI. Others have spent much more time, including this great strategy. I agree with a lot of it, but I put somewhat more weight on two things:
1. Partner with the US as the mainline strategy.
2. Europe's biggest economic opportunity may be in complements to frontier AI (not substituting): robotics, advanced manufacturing, pharma, biotech. Leverage existing strengths and ask what becomes dramatically more valuable if frontier AI exists?
More generally, a lot of AI strategy isn't really AI-specific. It's the familiar required reform agenda covering labor markets, capital markets, etc. All made considerably more urgent by AI.
"We currently have 5% of global AI compute ... that's just not enough."
@bakkermichiel on the immediate objectives for European AI strategy:
"We should secure access to the AI supply chain. This means chips and models ... we've seen that Europe having continued access to models is not a given."
"We have to get ready for ... this transformation"
"I've had policymakers in Brussels, in the Netherlands, elsewhere, reach out and ask: 'All this agent hacking, we thought this was years away--how should we actually get ready?'"
"Then there is building out our AI compute supply, to have leverage, to run models for Europeans, we need more AI compute."
Since this gained some traction, I have to frame this in what I think of it generally.
Europe desperately NEEDS incomparably stronger AI capabilities than it has right now. It should be a non-negotiable strategic goal.
Does it currently look like EU treats it as such? No. European Union had 3 years to start really pushing into this direction, and I'm not even asking for a good push, just a good attempt at it. All we've realistically see is it's being treated as a recurring talking point.
And it's just really sad for me. You will then see the same politicians talking even louder about the great brain drain and exodus of talented people. Guess why?
Come on Europe.
This is an insanely important report about how completely, utterly and totally unprepared Europe is for the arrival of transformative AI.
A single data centre in Malaysia will soon have a third of the AI computing power of *all* of Europe combined. Two data centres in the US will have more than our entire continent.
And that's just page one...
More than 60 of Europe's top minds (including two Nobel laureates) have just published "A Transformative AI Strategy for Europe", an independent report about how dire the situation is. As the world gets darker, we risk being squeezed and trampled by the US and China. On the current path, the report warns, even Europe's richest economies could slide toward middle-income status in a world run from Silicon Valley and Beijing.
Here's where we are right now:
→ The EU hosts just 5% of the world's AI compute. The US: 75%.
→ Not one of the ten most valuable AI companies in the world is European.
→ And this year, Europe got locked out of the world's most powerful AI model, twice, by American export rules. The EU's own cybersecurity agency had to negotiate for weeks to get in, and still doesn't have the latest version.
For years the comfortable story was: America innovates, Europe regulates. Well, that story is dead. You can't regulate what you can't access. And you can't protect 450 million people with second-rate cyberdefence against first-rate attackers.
The good news is that we have some leverage: Europe holds the single biggest chokepoint in the global chip supply chain (ASML). And this report sketches a plan to increase our leverage massively. Here's what we need to do:
1. Build a member state alliance for supply chain security – Netherlands, Germany and France at minimum – founded before the end of 2026.
2. Recruit many more frontier AI experts inside European governments ASAP.
3. Aim for at least 15% of global AI compute on European soil by 2030, built so local communities actually benefit.
4. Make Europe the world leader in AI assurance: hardware that can prove how AI is being used.
5. Prepare for crises and emergencies, such as engineered viruses and highly advanced automated cyberattacks.
It's too late to catch up with the US and China. But it's not too late to build leverage .
Read the full report here --> https://t.co/Byi5WYC8yW
Der Etwurf zur Steuerreform aus dem @BMF_Bund ist wichtig, aber das entscheidende Papier zur Wettbewerbsfähigkeit Deutschlands im KI-Zeitalter ist das hier. Wir brauchen Chips, RZ-Kapazität, hillige Energie. Danke @FuestClemens@ischinger et al. https://t.co/cNB1EwEs2G
As AI advances at a rapid pace, Europe faces a critical challenge: safeguarding its sovereignty and security. MSC Tech Strategy Initiative rotating member @privitera_ and his team of experts outline a path forward with a transformative AI strategy for Europe.
Read more 👇
An important report on how Europe can prepare for transformative AI.
I particularly welcome the proposals to strengthen technical expertise in European institutions, build resilience to AI risks and deepen international cooperation.
We can’t complain about Europe lagging in AI while dismissing every effort to catch up.
Surprise: no single initiative will fix everything. Progress has to start somewhere.
I'm increasingly bothered by people dismissing and joking about every effort as “not enough”. If you do that without contributing yourself or suggesting an improvement, then you're part of the problem.
Start supporting the people who are making an effort if you want something to change.
First, great initiative. I'm really happy it exists.
My main concern is that the proposal explicitly tries to reverse-engineer a lab based on the desired outcome (thousands of employees, multi-GW compute capacity, major acquisitions).
But reaching the frontier once isn't really the objective. What you want is an engine that can repeatedly produce great models with fast cycles. The first model will be bad, the second one a bit better, the third one will be decent, etc. You only get good at building models by iterating a lot.
To be fair, the proposal mentions the risk of overcommitting: it suggest independent technical reviews after ~18 months and then more frequently. But it also pre-commits GWs of compute in the first three years (with a goal of 16 GW).
I would structure the commitment differently. Political commitment (money, compute, etc.) is mandatory from day one, so the next election can't (easily) stop the project. Operationally, however, you probably want to start with 50-100 people, 100s of MW and start shipping. If that team can ship competitive models every few months, use its compute efficiently, attract users, and show that scaling is the bottleneck, then scaling toward 1 GW and beyond makes sense.
Decoupling political commitment and operational footprint matters because compute won't be a bottleneck at the beginning anyway. You'll need to build all the infra and learn to work together. It also derisks the most expensive part of the project. This leaves room to grow organically.
More specifically, I'd also assign 99% of the compute to research and training, 1% to inference. The reason is you need guaranteed access to compute to train models, but you don't necessarily need to serve them to millions of users. If you have good models, every inference provider will serve it for you anyway. You can always commit more compute to it later (good problem to have).
This is also why I think an open-weight strategy should be central to the project. Model releases are the best possible forcing function: fixed deadlines, external evaluation, real-world usage, and user feedback at a scale you can't reproduce with technical reviews. They push the organization to build models that people actually want to use instead of overfitting internal evals.
Even if you don't want to open-weight your frontier checkpoints, you can still release smaller models and previous generations by default. I really believe this should be essential and not an afterthought. The value you get from open-source adoption is really that insane.
Overall, this is a good start but the objective function feels slightly off. It should optimize for building the fastest possible learning loop. Compute, headcount, acquisitions, and infrastructure should only scale when there's evidence they'd make the loop faster.