Check against delivery

Speech by Alain Berset, Secretary General of the Council of Europe

OPENING

Minister Keldo,

Digitalswitzerland President Meyer,

OFCOM Director Maissen,

Ladies and Gentlemen,

It is a pleasure to join you today for this conversation on AI governance.

I. WHERE WE WERE

At Davos a year ago, there was a widely held view.

Europe was leading the race to regulate artificial intelligence.

The United States and China were leading the race to build it.

It suggested that speed and scale on one side could coexist with law and safeguards on the other.

That innovation and governance did not have to cancel each other out.

II. WHERE WE ARE NOW

One year on, the ground has shifted.

Implementation of Europe’s AI rules is being questioned and simplification debated.

Critics across Europe warn that what is framed as pragmatism risks weakening regulation itself and, with it, Europe’s capacity to set the rules.

At the same time, those who helped shape those rules are facing sanctions and direct geopolitical pressure.

III. THE REAL RISK

I am not saying regulation is collapsing.

The real risk is a race to the bottom.

A race where safeguards are treated as friction.

Where governance competes by lowering standards rather than defending them.

IV. WHY I AM HERE

I am here as Secretary General of the Council of Europe.

My responsibility is to ensure that artificial intelligence respects fundamental values:
democracy, human rights, and the rule of law.

These values are contested, including in Europe.

They are what endures when everything else breaks down.

V. WHAT WE ARE ALREADY SEEING

Over the past weeks and days alone, we have seen how AI is shaping events at the centre of unfolding crises.

Disinformation distorting protest movements.

Fabricated content filling the gaps left by censorship and blackouts.

Images circulating faster than verification can keep up.

VI. GEOPOLITICS & AI

In Iran, the protests are real.

The anger is real.

The repression is real.

And yet, almost immediately, facts are questioned.

In Venezuela, AI-generated videos claimed to show people celebrating events that never happened.

In Greenland, images appeared of children waving U.S. flags.

None of it was real.

VII. EVERYDAY HARM

And alongside geopolitics, we see something closer to everyday life — and deeply disturbing.

AI systems that remove consent.

That undress women and children digitally.

That normalise abuse at scale.

VIII. WHAT THAT MEANS FOR DEMOCRACY

These systems are increasingly reshaping how we think and learn.

Children use AI to do their homework.

Students use it to write their essays.

Entire populations begin to outsource judgment.

And we are already seeing the effects.

Literacy and critical thinking decline.

Instant reaction replaces reflection.

Clicks and “likes” replace debate.

Even our language is changing.

“Rage bait” was last year’s Oxford English Dictionary word of the year.

But outrage cannot be the business model of democracy.

AI must not be allowed to reproduce, amplify, and automate bias and discrimination.

And disinformation cannot be allowed to shape electoral choices, as recent elections across Europe have come under pressure.

IX. THE GENIE IS OUT OF THE BOTTLE

Today, the challenge is not getting information.

It is knowing what to believe.

The question is not whether AI will accelerate.

It will.

The genie is out of the bottle.

The question is: in a moment of global instability and uncertainty, will we let that acceleration pull us downward?

Or will we hold the line on the values that make innovation responsible and sustainable?

X. WHAT THE AI CONVENTION DOES

We are facing two paths.

One where technology protects, connects, and lifts people up.

Another where it divides, exploits, and erodes trust.

The Council of Europe’s Framework Convention on Artificial Intelligence draws that line.

It does not try to slow innovation.

It does not try to predict every future use of AI.

It takes a risk-based approach and offers practical tools.

And it is built not just for Europe, but for every country that believes AI must work for people — not over them.

XI. WHAT MAKES IT WORK

This is part of a longer journey.

Fifteen years ago, the Council of Europe adopted the Convention on Cybercrime, recognising early that digital power without rules would undermine trust and security.

In 2024, we did the same for artificial intelligence, by placing it explicitly under the requirements of human rights, democracy, and the rule of law.

The AI Convention is binding, but flexible.

It looks at AI across its entire lifecycle — not just at the end, since risks are often built in early.

It is about building rights into technology, transparency into systems, and confidence into public debate.

And it is designed to work across different legal systems, cultures, and capacities.

Since it opened for signature in September 2024, close to twenty states and the European Union have signed the AI Convention.

XII. FROM PRINCIPLES TO PRACTICE

A Convention alone is not enough.

That is why we have developed risk and impact assessments.

They help identify potential manipulation and interference at an early stage of deployment.

Switzerland shows how this works in practice.

Swiss Ambassador Thomas Schneider chaired the Committee on Artificial Intelligence that negotiated the Convention.

And Switzerland continues to promote a pragmatic, sector-specific approach that welcomes innovation while translating shared principles into national practice.

That approach is now being applied in close collaboration with the ETH Zurich ecosystem.

It supports collective intelligence, cultural and linguistic diversity, and informed choice.

In the end, this is not about having all the answers in advance, but learning early enough to correct course.

We look forward to Switzerland’s ratification of the AI Convention.

CONCLUSION

Stephen Hawking once said that powerful artificial intelligence could be either the best, or the worst thing, ever to happen to humanity.

This is a defining moment in the geopolitics of AI regulation.

The future we build will depend on governance.

If states do not shape AI, AI will shape states.

No country can do this alone.

And companies cannot stand aside.

We need to think in the long term and build to last.

So that the decisions we take today still hold in ten, twenty, or thirty years.

Thank you for your attention.