MEPs debate on the world’s first set of comprehensive rules to manage Artificial Intelligence's (AI) risks, and promote uses of AI that align with EU values, including human oversight, privacy, and non-discrimination. The rules follow a risk-based approach and establish obligations for providers and users depending on the level of risk that the technology can generate. They would prohibit AI systems with unacceptable levels of risk. Extracts from the debate.

digital agenda (T)

Strasbourg

debate

plenary session

use of robotics and artificial intelligence (T)

CLUNE, Deirdre (EPP, IE)

VESTAGER, Margrethe (DK, CE)

BENIFEI, Brando (S&D, IT)

VITANOV, Petar (S&D, BG)

VAN SPARRENTAK, Kim (Greens/EFA, NL)

TUDORACHE, Dragoş (Renew, RO)

Artificial Intelligence Act: extracts from the debate

Title:  Artificial Intelligence Act: extracts from the debate (I-242459)

SOUNDBITE (Italian), Brando BENIFEI (S&D, IT), Co-Rapporteur, Our position contains many improvements on the Commission's proposal, clarifying the text and bringing it up to date. Legal certainty Avoiding overlap with existing legislation. Stimulating innovation, but at the same time keeping human beings and their fundamental rights at the heart of all our endeavours. That must remain the centre point of Europe's approach to artificial intelligence.

SOUNDBITE (Italian), Brando BENIFEI (S&D, IT), Co-Rapporteur, Just to quote a few of the key points in our report, we have a clear definition in line with the work of international organisations, a chapter on Chat CPT that was originally admitted to avoid risks of discrimination and abuse of illegal content, then watermarking so that you can identify AI generated products

SOUNDBITE (English), Dragos TUDORACHE (Renew, RO), Co-Rapporteur, Providing through text developed in Parliament Start-ups and SMEs in Europe with a strong network of sandboxes so they can innovate, they can grow and develop technology in accordance with European values and in respect to fundamental rights, increasing the trust in their products as well as their competitiveness.

SOUNDBITE (English), Dragos TUDORACHE (Renew, RO), Co-Rapporteur, We have worked to put in place superheroes governance structure that makes sense. It is meant to ensure that we consolidate the digital single market, that member States have the proper tools and support in place for the enforcement of this regulation, that Europe has a futureproof approach to new developments in the AI and that we don't repeat some of the mistakes we've done with GDPR.

SOUNDBITE (English), Dragos TUDORACHE (Renew, RO), Co-Rapporteur, Ten or even five years down the line it is this governance structure that will give Europe the ability to deal with the rapid evolution of the AI and to reap the most benefits from it. And we have worked, first and foremost, to ensure our citizens rights and freedoms are not just respected but protected and strengthened. We don't want mass surveillance. We don't want social scoring. We don't want predictive policing in the European Union full stop.

SOUNDBITE (English), Margrethe VESTAGER, Executive Vice-President of the European Commission, When it comes to technology the most important thing when you embrace it is that you can trust it. And only by embracing something can you control it. If you keep it at arm's length, very difficult to get a tight grip. And what we can do here is to create trust, legal certainty, to enable AI to develop in a positive manner. To serve our society, our ambitions, the sectors, the businesses, the industry. We can do amazing thing with artificial intelligence when we can trust it.

SOUNDBITE (English), Margrethe VESTAGER, Executive Vice-President of the European Commission, When we started our journey, I frequently got the question, Why are you regulating A.I.? This will hamper innovation. Why not just let it go? But I think by now everyone realises that this is the important moment. Yes, this is the time.

SOUNDBITE (English), Deirdre CLUNE (EPP, IE), We're among the first to tackle the fast moving and evolving technology with a concrete legislative proposal on how to address those powerful foundation models to build trust and provide transparency and oversight of these systems. The absolute minimum that we need to offer here is transparency. It must be clear that this content has not been made by a human.

SOUNDBITE (English), Deirdre CLUNE (EPP, IE), And we also go one step further and ask developers of these large models to be more transparent and share their information with providers and how these systems were trained and how they were developed. This should address also the environmental sustainability of these systems.

SOUNDBITE (English), Petar VITANOV (S&D, BG), Colleagues from EPP, we will try to make the case how these systems could be used to catch criminals. But one thing is very clear to me: those who want to hide from our systems will always find a way to do so, and always the innocent will suffer.

SOUNDBITE (English), Petar VITANOV (S&D, BG), So it's not a question of whether the AI system have the potential to result in racially biased and discriminatory outcomes. We actually know for sure that this is the case. We see this confirmation of this in the data provided of multiple NGOs. We keep seeing it during the LIBE mission to Washington D.C., and we have heard it even from the EU High Commissioner for Human Rights. And no, AI is not dangerous only when used by autocratic governments.

SOUNDBITE (Dutch), Kim VAN SPARRENTAK (Greens/EFA, NL), I think that we should look to try to find the good in AI. That part should enables us to bring about progress for people in the environment while limiting the more detrimental sides. So after a year and a half and hundreds of hours of negotiations, I'm proud that the European Parliament is now starting to make choices.

SOUNDBITE (Dutch), Kim VAN SPARRENTAK (Greens/EFA, NL), We can vote tomorrow on a strong right to information and consultation so as to ensure that people won't be powerless in the face of computers and they'll know when algorithms are being used. A compulsory test for human rights before A.I. is used; strict regulations about how you generate and train systems and how you use data; and obviously this is particularly important when it comes to discriminatory data, and we also have to ensure that for major models such as Chat GPT have extremely strict rules.