Artificial intelligence: in Europe, innovation and safety go hand in hand | Statement by Commissioner Thierry Breton

Artificial Intelligence will revolutionise many aspects of our lives.

A bit like when cars arrived at the beginning of the last century. First we discovered that the car was a wonderful source of freedom, then accidents prompted society to regulate and road safety rules.

AI raises many questions – socially, ethically and economically.

But now is not the time to press the pause button.

The legal certainty start-ups need to innovate

On the contrary, the challenge is to act quickly and take responsibility for exploiting all the advances while controlling the risks. And it is regulation that gives start-ups the legal certainty they need to innovate.

As soon as I arrived at the Commission, I worked on regulation that would provide European start-ups and entrepreneurs with a framework that would enable them to innovate, while offering our fellow citizens the best possible security.

I conducted a wide-ranging consultation with all the stakeholders, from businesses to governments, NGOs to trade unions, researchers to start-ups.

Then, in April 2021, we presented our proposal for a European regulation, with the aim of  encouraging innovation and limitating risks.

The result is a balanced proposal that both promotes innovation and addresses safety imperatives.

What are the risks of AI?

The majority of applications using AI are risk-free. This is not the case for some. Social scoring, for example.

China uses it to control individuals. Using data captured without their knowledge, citizens are classified, evaluated and monitored.

We are proposing to ban this type of artificial intelligence application in Europe. Just like real-time facial recognition when it is used systematically.

But we also need to be able to use artificial intelligence, with appropriate safeguards, when it can provide us with solutions. This is the case, for example, in the field of health. It’s fantastic to be able to interpret an X-ray with the help of all the knowledge accumulated through access to thousands of cases of similar pathologies.

But we still need to know what data has been used, and who designed and manages the application. To be authorised in the European Union, this type of application will have to be certified before it can be used. A bit like approving a car model before it is put on the road.

Finally, other commonly used applications, such as those that generate texts, photos or voices artificially, like ChatGPT or MidJourney, will have to comply with transparency obligations.

A democratic process involving the 27 Member States, who agreed in December 2022, and the European Parliament in June 2023

So, for the last three years, we have been engaged in a democratic process to draft European regulation of AI.

The 27 Member States  all agreed to this proposal in December 2022, with a few amendments.

The European Parliament, for its part, voted on its position last Wednesday.

We are now entering the final phase of our democratic process to converge our final positions in the coming weeks, under the watchful eye of the Commission (what we call the “trilogues”, in Brussels jargon).

A clear legal framework always enables start-ups to get off the ground, guaranteeing them de facto legal certainty for their innovations

At the same time, this regulation must ensure that we can continue to innovate in this sector of the future, while protecting our fellow citizens from the potential risks involved.

To be able to create European leaders in AI.

Let me stress once again: a clear legal framework always enables start-ups to get off the ground, guaranteeing them de facto legal certainty for their innovations.

I call on the European Parliament and the Member States to adopt the AI law in the next few months.

We have no time to lose

We cannot afford to sit back and wait for the new law to become applicable in 2026. This is why I have started working with AI developers on an “AI Pact” to anticipate its implementation.

The AI Pact will bring together, on a voluntary basis, the main EU and non-EU actors in this field to inform and raise awareness of the principles and democratic process underlying the EU AI law. It will be fully aligned with the final text adopted by the EU co-legislators.

Next stop US (California), South Korea and Japan (G7 Tour)

I’ll be travelling to San Francisco next Thursday and Friday to meet the CEOs of Meta, Mark Zuckerberg, and OpenAI, Sam Altman, to discuss the AI Pact.

Following on from this, I will be travelling to South Korea and Japan to discuss the digital agenda with the Korean and Japanese governments, as they are keen to engage with us, particularly on AI, in line with the G7 statement on the importance of global coordination on artificial intelligence.

We are working together to make Europe, the world’s largest digital market, the continent of choice for AI start-ups, by promoting a clear legal framework that offers them the best guarantees of legal certainty for their innovations.

And safety for our 440 million fellow citizens. Because innovation and safety go hand in hand.