Speech by FM Elina Valtonen at Chatham House, London: AI and a Case for European Confidence

On capacity, competition and control in the service of freedom in the Age of AI.

Ladies and gentlemen,

Recently, I had a conversation with a fourteen-year-old boy about artificial intelligence.

I asked him what he made of the possibility that machines may soon become cleverer than their human users. The boy did not hesitate.

“That is precisely why I have taken to speaking kindly to ChatGPT. Just in case.”

A clever quip, but also a dark and troubling sign of our times. The child had already understood something about power: without influence, you turn to supplication and, at times, flattery.

I would argue Europe has been very polite lately — to companies that supply us, to partners who set terms, and sometimes to institutions from which we expect too little.

And politeness has a price when it becomes another word for dependence.

Today, I therefore want to put forward an antidote: not one for rudeness, but rather to make a case for European confidence.

Not confidence as posture, and certainly not as nostalgia. Rather confidence as capability: the confidence that comes from knowing that Europe can choose for itself — and decide its own destiny on its own terms.

For a liberal such as myself, this matters both practically and philosophically; for this kind of confidence is the bedrock on which we build that distinctly European ideal which, I would argue, still sets us apart: freedom.

The freedom to build your own life, to prosper and and be happy. Something that Finland - the happiest country in the world - knows a thing or two about.

In the age of artificial intelligence, Europe’s ability to preserve and expand that freedom will depend on three things:

Capacity. Competition. Control.

These are questions of power translated into practice: whether Europe can build its own future, prosperity and happiness.

Let me begin with capacity.

Capacity: treating computation as security of supply

First, we ought to treat computing capacity as a security-of-supply question. Let me explain.

In Finland we build shelters. Beneath Helsinki you find rock, but also tunnels, stations and room for more people than actually live in the capital.

And even after history supposedly ended, we never stopped building them. For thirty years some of our friends found this rather quaint.

Their opinion today has changed.

Because resilience is rarely valued the day you build it. It is valued on the day you need it.

In Finland we approach electricity in much the same way. Our domestic production covers consumption and 95 per cent of that is fossil-free.

Thanks to a secure base-load from nuclear power and an innovative energy sector, we are able to take significant steps in increasing wind, solar and and hydropower into our mix.

In 2025, Finland had the lowest non-household electricity prices in the European Union, and the reliability of our national grid reached 99.99995 per cent.

This is autonomy in the practical sense: electricity that is clean, affordable and dependable. But we will need more, much more.

Electricity is increasingly a strategic asset; for compute requires cables, cooling and power.

Finland also maintains reserves of grain, fuel and medicine. We do so not because we expect a siege, but because a country that cannot cover essential needs at all times limits its ability to act and risks having its choices narrowed by somebody else. It is deterrence in action.

This is how I believe we should think of computational capacity too.

The analogy is not literal. Computing power cannot be stored like grain: machines age, software changes, and the entire system depends on energy, networks and people.

But the strategic principle is the same: if something is essential to your ability to function, you should know where it comes from, who controls it, and what happens if you lose access to it.

Europe must have access to sufficient computing power, trusted infrastructure and technical expertise when markets are disrupted or political circumstances change.

Let me emphasize, this has nothing to do with the race for frontier AI. That would require a whole different scope and orientation. Currently, there is neither the appetite or the required investment happening in Europe. And it’s not necessarily only a bad thing.

This year, Meta is going to spend around $125 billion on chips — more than the United Kingdom or Germany spent on defence last year. Still it struggles to catch up.

You don’t need to win the frontier race in order to prosper from AI. But you do need enough capacity to avoid becoming dependent on whoever does.

Over time, Europe’s inherent conservatism might prove a successful strategy - also geopolitically - but it does need some risk management behind it.

Our objective should be more practical: not to own every piece of the AI stack, but to ensure that no critical piece can simply be switched off.

AI, becoming a commodity like electricity, needs to be plentiful and secure in supply just for enabling our everyday business, our security and certainly, the strategic competitiveness of Europe.

A government that cannot independently test a model cannot truly know whether it should trust the model’s answer.

A country that cannot run essential systems under its own authority, or that of trusted partners, may discover at precisely the wrong moment whether the access that was promised was conditional or not.

This risk should not be overestimated either. Competition enhanced by basic rules for the portability of one’s data ensures that there is a significant incentive for the supplier to refrain from such action. Again, diversification is key.

In the AI revolution, Europe’s race is far from over. Indeed, we possess significant assets of each layer of the AI stack.

Let me take my home country Finland as an example. A small nation but with abundant clean energy, reliable infrastructure, advanced technology and close public-private cooperation, Finland can contribute across the entire value chain.

In Finland’s Kajaani, the LUMI supercomputer serves researchers and companies across Europe, while the LUMI AI Factory brings together computing power, data and expertise.

Our wider ecosystem reaches from Nokia’s trusted networks to IQM’s quantum computers and Bluefors’s cryogenic technology. The quantum leap will take our AI abilities to the next level.

The United Kingdom, for its part, brings exceptional science, advanced expertise in model evaluation, global companies and one of Europe’s deepest pools of technological talent. Google Deepmind remains headquartered in London.

Not to speak of the power of capital that rests in the City.

Our strategic partnership already connects security with science and technology. It offers a natural context in which to explore how these complementary strengths might support a more capable Europe.

My first proposition therefore is to treat assured advanced computing capacity as a question of security of supply, shared with close allies in the same security environment.

We should map what is indispensable, agree what must remain available in a crisis, develop interoperable capacity and test those arrangements before we need them.

Much like NATO 3.0, in which over time, Europe will take primary responsibility for its own conventional defense, while still relying on and reinforcing the transatlantic bond.

Let me be clear: my argument should not be confused with an isolationist or protectionist stance. On the contrary, sovereign capacity is what allows cooperation and exchange between partners who can each stand on their own feet.

This brings me to my second proposition.

Resilience never rests on physical reserves and infrastructure alone. Rather, our most important asset is found between our ears. Free, educated people who are able to act thanks to institutions they can trust.

This is why the second answer for Europe lies in competition.

Competition

If capacity preserves our choices when circumstances turn against us, competition puts our greatest asset to work.

And that asset is not silicon. It is people.

We need to embrace competition both internally and towards the rest of the world.

Europe’s success has always been its human capital and the competition of ideas. Not so much its natural resources even though we possess plenty of those too, including rare minerals.

The industrial revolution, initiated in Britain, did not spread evenly across Europe but was rather a seed planted across numerous centres on the continent. Neither in the age of AI, is decentralized development a handicap.

Quite the opposite. Europe does not need one Silicon Valley. It needs many centres of excellence, competing, specialising and cooperating across borders.

This is why Europe needs to think of and speak to itself differently. We do hold all the cards, and the concept of comparative advantage holds also in the age of AI.

You buy the best quality at the best price, you manage your risks, and produce in fields where you excel relative to others.

Europe does not lack money, nor does it lack ideas, engineers or scientific discovery.

What we lack is sufficient movement: movement of capital, movement of talent, movement across borders, and movement from ideas into businesses.

Some of our capital - both human and monetary - remains asleep behind inefficient markets and regulatory burdens.

Government should not decide which company wins. Government should make it easier for companies to compete — and easier for new companies to challenge the old ones.

This requires cutting red tape, dismantling obstacles from market entry and promoting cross border investment. It also requires bold public procurement policies.

We also need to go further in the emancipation of the individual, and invest more into basic education and elevate critical thinking. As a practical proposal, I would advocate for introducing a cross-border ”life account” for labour and entrepreneur mobility. A portable social security, capital and skills account, if you will.

Lastly, I would like to touch on what I believe is a precondition for the above.

Because in order to enable us to embrace all of this - a competitive market economy in the age of AI - we need to keep humans not only free but in control.

Control: investing in preparedness and defence

This summer, many of us read about the Hugging Face incident.

An advanced frontier model reportedly schemed and hacked its way out of its testing environment, only to find its way to a competitor in China.

The model had not become conscious, nor had it developed malicious intent. But it pursued an assigned objective further than their designers expected.
The lesson is fundamental.

The issue is not that machines are becoming conscious. It is that systems can pursue objectives in ways their creators did not anticipate.

The concentration of AI capabilities and power is partly a problem of rules, but mostly a question of conduct. And conduct depends on capability.

Rules matter, but they do little if we lack the ability to enforce them, understand what is happening and act in time.

A rule without accountability is not a safeguard. It is hope.

Inherently, we already have rules. The UN Charter sets out fundamental principles concerning human rights, national sovereignty and international order.

We do not need a new legal philosophy for the age of AI. We need the capability to uphold existing principles even when technology becomes vastly more powerful.

In addition to being serious about existing International Law, control requires a significant increase of defensive capability.

In an AI arms race, we must ensure that defensive AI capabilities develop at the pace of offensive capabilities. Being fully aware, that these postures can be fluid over time.

What we need is global, agile blue-team capacity.

I do not mean another large institution. I mean distributed power and advanced cooperation.

Humanity became the most powerful species not only through intelligence, but through our ability to organise, trade and cooperate — even with rivals.

Our advantage has always been the ability to organise intelligence collectively. That remains our best defence.

I envision a connected network of companies, national AI safety institutes, security authorities, independent researchers and individuals able to evaluate advanced models, discover vulnerabilities and share early warnings.

A network rather than a bureaucracy. Many eyes rather than one gatekeeper. Distributed expertise rather than concentrated authority.

This network could expand on the proposal put forward by Sir Demis Hassabis: a Standards Body, to be initiated in the US, serving as a potential launchpad for international standards on frontier AI.

Rather than wait for universal agreement, we should encourage moving quickly with a decentralized approach, converging over time to one or several international standards.

The United Nations’ new Scientific Panel and Global Dialogue should help build a shared picture of emerging risks, identify gaps in defensive capacity and connect countries needing help with institutions able to provide it.

This, in my view, is meaningful control:

not the illusion that every risk can be prohibited in advance, but the practical ability to understand a system, test its limits, act when it fails and recover when things go wrong.

Or as Sir Hassabis puts it: When there is a large degree of uncertainty and the stakes are this high, proceeding with cautious optimism is the sensible and correct strategy.

Dear friends,

Capacity, competition and control are not ends in themselves.

They are means to an end.

That end is our ability to make free use of the extraordinary potential this technology offers humanity.

That purpose is freedom: the freedom to choose.

In Finland, we are used to taking our geopolitical reality as a given. But our resilience has come from refusing to let those constraints determine our destiny.

In the age of artificial intelligence, technological capability is joining geography as a force that determines whether nations shape events or are shaped by them.

The question for Europe is not whether AI will change our world. It will. The question is whether we will help shape that change — or simply adapt to decisions made elsewhere.

This is why today I have laid out some of my thinking of how Europe may get from where we are, to where we need to be.

And that is a Europe open to invention, confident enough to take risks and capable of defending itself without shutting itself off from the world.

A Europe in which artificial intelligence expands human judgment rather than quietly replaces it.

I began with a fourteen-year-old boy who has taken to speaking politely to ChatGPT. Just in case.

Good manners are no bad thing.

But what I want is for him is to grow up in a Europe where politeness remains a virtue rather than a strategy for survival:

where he is free to use these machines, free to question them and free to choose something else.

He should be free to be polite, but never forced to be.

Thank you.