MEPs vote on the world’s first set of comprehensive rules to manage AI risk and promote uses of AI that align with EU values including human oversight, privacy, and non-discrimination. The vote was adopted.

statement

plenary session

vote

METSOLA, Roberta (EPP, MT, EP President)

use of robotics and artificial intelligence (T)

BENIFEI, Brando (S&D, IT)

TUDORACHE, Dragoş (Renew, RO)

Artificial Intelligence Act: extracts from the vote and statement by Brando BENIFEI (S&D, IT) and Dragoş TUDORACHE (Renew Europe, RO), Rapporteurs

Roberta METSOLA, EP President announces the opening of the vote and members voting, 4 shots

Roberta METSOLA, EP President announces that the vote is adopted and congratulates, 449 in favour, 28against, 93 abstentions. 3 shots.

Dragos TUDORACHE, requests sending the report back to inter-interinstitutional negotiations and the request is adopted, 5 shots

SOUNDBITE (English), Drago&#351; TUDORACHE (Renew, RO), Co-rapporteur, It's important to have rules because what we had until now was just principles of self discipline on the part of the company. However, the company is developing and we saw that it is not enough. Here in Europe, we believe that is not enough and we believe that some standards or some guardrails are important to protect the interests and the values that we hold near the union. And that's precisely what the legislation does, and it doesn't do it. And I think that's important to stress. It doesn't do it by unnecessarily putting barriers on businesses and for for creativity and innovation, because it limits very much the scope of those that will have to comply only for high risk applications.

SOUNDBITE (English), Dragoş TUDORACHE (Renew, RO), Co-rapporteur, We have decided not to regulate the technology itself. I think this would have been wrong, but to look at the uses of technology and then I will attempt was to categorise those uses that would bring about risks to the values of the interest that we want to protect. And we have created categories of risk. On top of these categories. Are those uses a way that is so harmful, that are so detrimental to the interests and to the values that protect us. We simply do not want them on the market. So we have introduced them in the list of prohibited practices.

SOUNDBITE (English), Drago&#351; TUDORACHE (Renew, RO), Co-rapporteur, I'll give one example social theory of the time that China does on a massive scale for its citizens. We say something like this we do not want for our citizens. We don't want to be pigeonholed and scores allocated to them on the basis of which they can access something certain services in society. That's not how we run our democracies. Another example, which was also subject to a lot of emotion for the vote today, which is biometric recognition in public spaces in real time. We as Parliament have said that, again, this is a use of technology that can deliver policy in our society. Same thing for emotion recognition of a predictive policing.

SOUNDBITE (English), Dragoş TUDORACHE (Renew, RO), Co-rapporteur, It is already proven that prompted in the wrong way, Chat GPT can generate content that can tell children how to harm themselves. Is it something that we want in our society, unchecked, either verified, uncontrolled, without some clarity as to the rules or responsibilities for that sort of content? My answer and our answer in this House is no.

SOUNDBITE (English), Brando BENIFEI (S&amp;D, IT), Co-rapporteur, We need to fight new types of disinformation. A.I. can easily generate, for example, deep fakes that can be easily spread. So we need to put a clear marking that a content is not the reality that what is created by A.I. should be identifiable and also the deep fakes kind of so where you have people saying doing things they have never done, this should be immediately recognisable because it's true that we need to educate people. It's true that we need to build a conscious of finding different sources, getting information in a more conscious way, but that's not sufficient.

SOUNDBITE (English), Brando BENIFEI (S&amp;D, IT), Co-rapporteur, We made a clear agreement and balanced text that are the large support in the Parliament that allows the use of facial recognition, biometric identification, not in real time with a judicial authorisation and for a crime that has been is under investigation and prosecution. This means that we can pursue criminals and cases that need to be pursued in terms of of risks of security for citizens, but without a real time biometric identification in public spaces that can be used for mass surveillance because it would be used necessarily to look at every person. And with that a larger use which can lead to a lot of problems.

SOUNDBITE (English), Brando BENIFEI (S&amp;D, IT), Co-rapporteur, They will need to show where they got what they got in terms of copyrighted material so that to train their eyes. So in this way we can give some support to auteurs, to creatives that need to know when creative material that they produced was used to train their eyes so that they can check if that copyright law was respected or if it was violated.

SOUNDBITE (English), Brando BENIFEI (S&amp;D, IT), Co-rapporteur, We hope we can have this regulation approved by the beginning of next year. This means then there will be two years time for full entry into force.