The European Parliament adopted in October 2020 three resolutions on AI covering ethics, civil liability, and intellectual property (IP).The legislative resolution by Axel Voss (EPP, DE) calls for harmonising the legal framework for civil liability claims and for imposing a regime of strict liability on operators of high-risk AI systems.

press conference

VOSS, Axel (EPP, DE)

use of robotics and artificial intelligence (T)

Press conference by Axel VOSS (EPP, DE), rapporteur on the presentation of JURI position on AI Act

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, Artificial intelligence, the A.I. Act is of great strategic relevance. AI is the booster into digitalization, and that's why we should really take the time to ensure the rules are practicable and looked at with that end. There are various problems occurring that we must discuss. First of all, about what high risk oriented is to try and catch that in the various rules, so that we don't need to be concerned that machines will be deciding something for us, and that is what it's primarily about.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, We must also consult today 90% of applications of a AI are not ones of concern constitutionally, but we will increasingly get there with further developments. Let me just make a few points of principle initially. In jury we must ensure that we can strike a balance between what we would like to regulate, what we would like to attempt to get under control in the usage of AI.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, But, on the other hand, not act so as to be an impediment to innovation. We want to ensure that the appropriate principles ethically or in terms of fundamental rights are fostered. But it all must be legally certain and practicable, in particular so that we can be sharing in those developments rather than running running along behind them, because it's extremely important for us that we are part of the development of A.I. in Europe.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, A couple of important points with respect to the things that we should keep an eye on and which possibly may be discussed from different viewpoints. Firstly, we now have a large number of digital acts, or acts on digital digital topics, to put it another way round. And we need to ensure that we do not overlap. We don't want to say on the one hand, something different to what we say on the other hand. It all must be brought together harmoniously within a whole.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, Another big point will be how scrutiny takes place. Here, of course, for artificial intelligence, there's a wish either to have an agency, a new European agency, or to use preexisting structures and reinforce them and improve them. With that in mind, we'll have a discussion about how we do that as a whole.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, My concern going above and beyond that is that in this area, it shouldn't just be artificial intelligence, but also the Digital Market Act and the Fundamental Rights Act that come into play. We need the oversight of the whole. We shouldn't have things decided in complete independence from the other. It all has to be part of an overall approach, looking at how we come alongside digitalization.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, We'll also have to resolve a controversy arising from the fact that we have the GDPR and that we, as the legislator, coming up with a result allowing us to work with AI that touches on our ethical concerns. We want to have unprejudiced A.I., or gender balance A.I and we don't want any discrimination in A.I.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, But these outcomes will only be achieved if the A.I. system is trained sufficiently. In other words, you need a great deal of data and high quality data in order to be able to achieve these results. And if we don't have them, then that development is stopped in its tracks, which means that person-related data are needed to train the system. And there you have the way in which the GDPR and the A.I. Act interact, and the sandbox is possibly a way of resolving that controversy, creating a space where, for the purposes of training, personal data can be involved without them, leading to a breach of the privacy or data protection of individuals. And that would be a potential resolution of that situation.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, A very decisive field, of course, is the question of what is a so-called high risk system when it comes to algorithms. And here I believe that we should focus more on what is really of concern to us, namely machine learning algorithms. The others are fairly static and not much would result. But machine learning is what we really need to focus on.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, Why are not all A.I. systems comprised? Well, perhaps only 10% could really be dangerous. It may even be less than 10%. And also, the proposal, at least in the A.I. Act, is so huge when there's a high risk application that an enormous number of criteria would need to be complied with in order to be able to use that A.I. system. There are conditions not just linked to the A.I. itself, but also on to those who are developing it or using it or distributing it.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, Which means that we need to ensure that this is reined in. Otherwise, the most simple applications involving A.I. will be affected. We need to come up with a decent version. For me, what it comes down to when questions are asked about what is a high risk system, we have to ask the question of what risk do we want to cover at all? Of course, you could always say straight off that fundamental rights are a regulatory level to cover, but when you look at that in detail, it just expands immensely, infinitely.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, And then, of course, when you think about liability linked to A.I., you've got life, the body, health, property. These might be risks that you want to cover. But here, we have to ask what risk do we actually wish to protect? To my mind, that means that you've got to look at the application case. Is it a critical case of application for the individual? Does that come down to critical use?

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, And the other is the environment of the application. The critical area where we have to ensure that, if both conditions are met, we are in a high risk area. Expressed differently, I want to exclude A.I. in a high-risk structure such as critical infrastructure not necessarily being labeled high risk. It may not collect high risk analytical data simply because it's in a high risk environment infrastructure.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, So that's the view I would have of the points of dispute or consideration. Of course, biometric data are something that we're going to have to get to grips with, in particular, because in the proposal that was laid down, and let me say that by a metric, data is protected by a corresponding regulation. We don't need to implement that protection here again. We just have to ensure that we have a sensible system that regulates it all.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, Moreover, mass surveillance is often a buzz word that's raised when it comes to facial recognition and such like, but here development has moved on quite away. There are privacy tech systems that have the possibility of being used for dangerous applications, but they should be implemented with sufficient checks and balances that individuals' data are not adversely affected.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, There are possibilities here. They just need to be authorized and understood, so that it all works that way. In broad terms, I would say it boils down to the fact that risk applications are scrutinized very closely and looked at from the perspective of governance while we focus on what is high risk. We need a trustworthy A.I. concept.

SOUNDBITE (German), Axel VOSS (EPP, DE), Rapporteur, That's from what the Commission has proposed and we must intent to ensure that our development is not impeded by overregulation. To my mind, we have to be part of the game developing a A.I. so that we have an opportunity to survive digitally. That's what it is all about at the end of the day.

Q&A Question regarding whether the high risk application with the use and the environment is supported by the other political groups; the liability and whether the Comission supports this issues in the proposal and the calendar for the adoption of the regulations.