Committee on the Internal Market and Consumer Protection Extraordinary meeting
Legend
Transcript markers
- text Sentence being played
- text Search result
- text Intervention cited by the link you opened
Click a sentence to play the video from that moment. Select text to quote it with the speaker, timestamp and link.
More options
Offset
Transcript offset
Export
Live capture
Versions of this video
Transcripts may differ slightly from one version to another.
IMCO European Parliament
Follow its sittingsDirk Gotink
This will stay. I mean President Metsola also was very happy with the that how we sort of made progress on the previous AMLA experience. So let's see. It's up to the next ones to make progress even more. Thank you very much and let's just vote it and get on with our agenda.
Thank you.
Thank you very much also to all of your work as rapporteurs. So with that I open the vote. There's one vote by roll-call vote. I open the vote.
Anna Cavazzini
I hope everything is working because I didn't do a test vote. Yours is not working? Can you use... yeah? Al de vez, sí. Yeah, it's late, no problem.
Okay, everyone can vote? Yeah, and I close the vote. Adopted with 38 in favor, 2 minus and 5 abstentions. So congratulations. Amazing.
So we will have the plenary vote, as I said, on Wednesday. This concludes agenda item 3. We now come to agenda item 5, a human-centered approach to AI governance. We put on the agenda on the request of several political groups, but mainly the EPP, the exchange of views with Professor Paolo Benanti on the encyclical letter Magnifica Hominatus of His Holiness Pope Leo XIV on safeguarding the human person in the time of AI.
And I think the coordinators decided to put this on the agenda because the discussions on AI and also the ethical dimensions of AI are getting bigger and bigger. If you check the news, there's even more and more companies themselves that ask for like a stop of the development because they fear AI might become, yeah, stronger and more intelligent and autonomous compared to the human beings. So we thought it's very interesting what has been written down in this encyclical. And I welcome again Professor Paolo Benanti who presents the encyclical for 10 minutes. And after that we have a round of members for asking questions.
So Professor, the floor is yours.
Paolo Benanti
Thank you very much. And as a European, I'm really glad for the work that you are doing and for how you are shaping the digital coexistence of the European in this time. Well, let me stitch to the plan and let me try to present the encyclical that is a broader and big document, trying to analyze the most interesting point for what we can define as a governance of technology.
The first point is what is the perspective of such kind of document. Well, it's not a policy document in a strictly way, so there is not a way that is suggested to be adopted. Is neither a philosophical document in the sense that is simple a reflection on the human condition or technological condition. Is something different, is something that belong to a long tradition of reflection inside the Catholic Church that is called social teaching. What does it mean?
Well, according to the core belief of the Christian faith, the commitment of the Christians in the history, in the time has to make visible some kind of principle of justice and peace. So, everything that you can find inside the social doctrine of the church should be a commitment of people to build with other people, not faithful, with all the people, justice and peace. And this kind of teaching arise as a corpus of teaching in itself with the industrial revolution and with the simple social transformation that industrial revolution put in itself in society.
That happened more than 100 years ago with a document that the name was Rerum Novarum and was made by another Pope that the name was Leo, Leo XIII. The problem was not the steam power at that time, but the steam power, the concentration of resource and the way in they was pinging on society was acting as a form of order and displacement of power. And the reaction of the church was about how to make it possible a coexistent of this new form of power inside the society. 150 years later with Magnifica Humanitas, the object is not artificial intelligence in itself like was not the steam power at that time, but is to recognize that such kind of technology is a form of power, is a form of force that could simple alterate the democratic and normal functioning of society. Defining in that perspective what is the goal of the document require me to simple clarify at one point why is such possible a things like ethics of technology.
And also to remove some possibility of misunderstanding connected to ethics of technology that could simple let's see a technological tool as something that is neutral or something that has not ethical dimension in itself. Well, in itself this is true, there is not an ethical dimension inside a technological artifact. But what is also true is that every time a technology is impacting a society, it works as a form of order and displacement of power. Let me be really clear. If we build a subway in a city, the place where we put the track and the place where we put the station are simple allowing someone to get the transportation and is denied to someone else to take the same transportation.
Ethics of technology in this sense is means questioning technology so the different stakeholder can simple see the difference and the transformation in society and how this difference and transformation in some way can be driven by the research of justice and the research of subsidiarity principle and other principle that are connected to democratic coexistence.
So if we look Magnificum Humanitas, the first third of the document is a sort of historical résumé of this kind of approach to technology that the church developed in the last 150 years. It's interesting, and it's interesting for me to present to you that represent an expression of the popular willing in European Union, because it's clearly stated in this 150 year of reflection that what is interested the church is not a Catholic definition of ethics of technology, but is a commitment of the church in itself to cooperate to anyone, not depending on the belief or faith, that are simple would like to produce an action to have a social justice in the public sphere. So it's not a research of a denomination or definition of one faith, but is open up to a broader discussion, a plural discussion, to try to put some kind of guardrail around such kind of technology to shape the democratic space. Going deeper inside the different part of the document before my 10 minutes will be consumed, I would like to stress at least four topics of interest that arise from the document.
The first topics is connected about the concentration of the market that is coming with this wave of technology. Such kind of concentration and such kind of capabilities to operate in the public space with this technology is really, really high, it's really, really concentrated in few hands and with a huge moat driven by capital, as it was not before in the history of technological innovation. And such kind of concentration of power and economical power is also another element that describe this challenging impact of this new society in the public space in a way that need to be politically addressed and ethically discussed to avoid some kind of threat or damages to the human dignity and to the dignity of every citizen. The second element is this ability of this technology to surrogate human decision in a way that can simple produce or reproduce a sort of choices that are not connected to the responsibility of a human decision. This bring back to two elements.
One elements we can summarize simple with something like, for example, defining problem like, you know, the one that Hannah Arendt bring on the table when fascism and nazism take power in Europe. And the problem was how this was possible.
Well the answer was the absence of any kind of friction inside society and this absence of the middle layer inside society produced a huge power on this kind of system. Well you can find a trace in all the document that say that if we remove another layer of friction that is the human decision for algorithmic decision making, well such kind of effect could be amplified or we could witness such more despotic effect in a decision making system that simple do not request anymore any kind of responsibility in the execution of the process. There is a third element that it's really visible now for example in some kind of experiment that are ongoing on asylum seeker and refugees. The tendency to have some kind of bias on human judgment and to think that an algorithmical judgment could be more fair and more neutral. In the asylum seeker cases there are a lot of situation in which we are removing the traditional way in which an officer is listening the history of the asylum seeker for an algorithmical decision making process that simple request to this human beings give me your data and I will tell you if you can access or not access to asylum seeking system.
Well, this new idea of judgment of decision-making system on the naked human beings, an asylum seeker is someone that has not anymore right from the country where he's coming and has not yet right in the country in which he's knocking the door, well, is the level that we have to understand as the level that we give to the human dignity in the public sphere. And the fourth one, and then I think that my 10 minutes are finished, is connected to what we can define an hacking of human culture. So probably we don't have and we'll never have a general artificial intelligence, but what we can have is a super manipulation system that can simple alterate the way in which we perceive information and the way in which we build our personal conviction on society, on ourselves and on other people.
Connected to the recent wave of apocalyptical thinking that some kind of company are pushing on the table connected to technological development, there is a section inside the document that recognize that the capital power, the technological power, has another form of power that is connected with them, that is ideological power, that sometime is using also religious expression like the apocalyptical one to simple convince of a sense of urgency that produce a decision in one direction or in another direction according to the convenience of the player that are producing such kind of discussion, but neither else that a political ideology that would like to take place now. And so the frontier of what need to be addressed to preserve the human dignity is this concentration of technology, the immense amount of money that are behind such kind of innovation, and the ideological factor that now are able to orientate also the big player in the public space. Well, I stop my short presentation here and I'm really open to any kind of question that you can have on the subject. Thank you.
Thank you so much, Professor. I mean, it's exactly also the debates we are having and had having here in the European Parliament, and that's why also we brought forward, I think, the first AI law on that scale in the world. Have already a lot of people asking for the floor. We start with the EPP, Pablo Arias Echeverría.
Pablo Arias Echeverría
Muchas gracias, presidenta. I'm going to speak in Spanish. Muchas gracias, profesor Benanti, por su exposición.
La verdad es que exponer en diez minutos lo que significa esta encíclica, pues la verdad es que es casi casi como escribir un tuit y hacer un resumen del Quijote en un tuit. Pero dicho esto, es verdad que le agradezco mucho porque yo creo que la encíclica nos ha dado mucha luz o nos ha arrojado mucha luz sobre lo que es y lo que puede llegar a ser la inteligencia artificial. Y desde esa, desde la perspectiva no solo, por supuesto, cristiana, pero sí desde una perspectiva de los valores y los principios, sí que se me antoja que es una herramienta que nos puede aportar grandes beneficios en cuanto a salud, en cuanto a soluciones a problemas complejos como pueda ser una enfermedad sin solución como un cáncer o de cualquier otra índole o en el ámbito empresarial, industrial, de cualquier cuestión. Pero también no está exento de riesgos, todo lo contrario. Y bueno, pues si bien a través de la religión pues conseguimos poner diez reglas muy simples como fueron los diez mandamientos, pues resulta que de ahí hemos conseguido tener una regulación basada en cuestiones que en todos los credos tienen esa base sobre la diferenciación entre lo que es el bien y el mal y sobre la diferenciación entre lo que se puede y no se puede hacer.
Yo no sé si esto es algo que se pueda plantear desde la perspectiva del desarrollo de la inteligencia artificial, pero sí que es verdad que esta encíclica ha puesto encima de la mesa eso que todos demandábamos y que probablemente no éramos capaces de profundizar para llegar a tener soluciones o respuestas a esto, ¿no? El poder tecnológico pues tiene poder manipulador, tiene poder de controlar. ¿El algoritmo puede reemplazar al ser humano? La pregunta es muy simple, muchos dicen que sí, muchos dicen que no. Yo tengo mis dudas al respecto, creo que puede ser muy útil, insisto, pero también creo que puede ser altamente destructivo.
Y lo que me gustaría saber es cómo podemos entre todos, a través de cuestiones como la encíclica que ha publicado el Papa, hacer de la inteligencia artificial algo realmente positivo para la humanidad y no simplemente para algunos, como ya hemos visto en el ámbito digital, que se pueden lucrar sobre ello, manipular e incluso poner en riesgo nuestro sistema de vida y nuestros modelos de democracia. Muchas gracias.
Laura Ballarín Cereza
Muchas gracias, presidenta. Voy a seguir en castellano y darle las gracias al profesor Benanti por su explicación, por aceptar también hablar a esta comisión y es un placer escucharle, sobre todo en un momento tan relevante en el que, como ha dicho la presidenta, los CEOs de las grandes empresas tecnológicas que están desarrollando esta inteligencia artificial, pues han puesto encima de la mesa los últimos días la necesidad de abrir un debate y saber si estamos avanzando demasiado rápido, si vamos a ser capaces de controlar esta inteligencia y, sobre todo, pues la necesidad de gobernanza. Yo creo que la encíclica, incluso para los ateos, fue un texto muy inspirador y creo que para este Parlamento también, sobre todo porque va en una línea muy parecida a la Ley de Inteligencia Artificial que desarrollamos en Europa, la primera en el mundo, sobre todo con el concepto de que esta tecnología puede ser poderosísima si se desarrolla bien, pero por supuesto que nunca debe convertirse en un sustituto de la dignidad humana y que siempre tiene que tener a las personas en el centro, evitando una concentración de poder peligrosísima. Entonces, yo voy a hacerle dos preguntas. Sabiendo que el siguiente modelo de inteligencia artificial está siendo desarrollado por la propia inteligencia artificial, ¿estamos llegando en un punto en el que estamos vamos tarde si queremos gobernarla, si queremos controlarla, porque es la propia IA la que está desarrollando este modelo?
Y en segundo lugar, ha mencionado el uso ideológico de la palabra apocalipsis y quería preguntarle si podemos confiar en estos grandes magnates tecnológicos cuando ahora nos advierten de los riesgos de esta inteligencia artificial, pero si a lo mejor esconden precisamente sus intereses, un interés espurio en que no la gobernemos, en evitar que se apliquen las leyes y cómo bajo estos estos términos apocalípticos a lo mejor se esconden intereses personales. Muchísimas gracias.
Esto por un lado.
Elisabeth Dieringer
Herzlichen Dank. Ja, ich möchte beginnen mit einer ersten Frage und zwar die bezieht sich auf diese Konsultation, die ja stark eingeflossen ist. Aber sind wir nicht an einem Punkt angelangt, wo wir praktisch die gesamte Spitzenexpertise auf der technischen Seite der KI von den ganz großen Unternehmen schon aufgesogen und angenommen wurde? Und das Schreiben selbst warnt ja vor einer immer weiter wachsenden Machtkonzentration. Und daher meine Frage, sind wir an dem Punkt angelangt, an dem das beste Wissen und das tiefste Verständnis dieser Systeme genau von jenen Unternehmen vereinnahmt worden ist, die diese Macht konzentrieren, sodass unvoreingenommene Informationen von der Forschungsfront von Außenstehenden schlicht nicht mehr zugänglich sind?
Und wie war das während Ihres Entstehungsprozesses? Wer blieb Ihnen als Ansprechpartner, der wirklich unabhängig von Ihnen war, denn die Kirche befindet sich ja doch auch in einer sehr privilegierten Position, in einer Position, die nicht unbedingt abhängig ist von den Fördermitteln oder von Investitionen? Und eine zweite Frage ist, und da möchte ich aus dem Schreiben auch zitieren, und zwar es wird der Unterschied zwischen einem Algorithmus und einer Person über den Umgang mit dem Fehler erklärt. Ich zitiere also wörtlich: Für einen Algorithmus ist ein Fehler ein zu behebender Mangel, für den Menschen hingegen kann ein Fehler Anstoß zu tiefgreifender Veränderung sein. Interessant finde ich, dass diese Systeme gerade dadurch funktionieren, dass sie ihre eigenen Fehler ausnutzen.
Das Modell einer Vorhersage, sie ist falsch, es misst, wie falsch sie war, minimiert diesen Wert, und genau aus diesem Fehler entsteht die nächste bessere Vorhersage. Es ist eine Art Fehlerlernmaschine.
Angesichts dieser Ähnlichkeit im Umgang mit Fehlern möchte ich daher fragen, ist der Algorithmus am Ende vielleicht doch irgendwie menschlich? Danke schön.
Stefano Cavedagna
Thank you chair, I would like to speak in Italian. Grazie presidente, grazie per per questa relazione, ci tengo a fare un'osservazione e a porre un un quesito. Abbiamo visto che l'intelligenza artificiale per sua natura e anche per suo sviluppo ha ovviamente, purtroppo aggiungerei, una tendenza all'efficienza, quindi di conseguenza non considerare quello che noi invece abbiamo sempre considerato il fattore umano o, se meglio preferisce presidente, anche il fattore divino, sacro, spirituale dell'individuo stesso. Nel momento in cui l'intelligenza artificiale deve porti una soluzione anche di impresa, aziendale, di organizzazione della società, probabilmente ti direbbe di far perdere il lavoro a determinate persone, ti consiglierebbe di eliminare degli individui che considererebbe un peso per la società mentre noi che conservatori riteniamo che invece ogni individuo ha una sua anima e una sua pari dignità non non si potrebbe accettare l'ipotesi di vedere messi su più piani differenti individui magari che sono meno produttivi o che vivono una particolare condizione di vita o una forma di patologia che non sono solo patologie fisiche ma quelle che vediamo oggi sono sempre più psicocognitive.
E quindi la domanda che io volevo fare e anche l'invito che volevo fare è cercare di sviluppare sistemi di tutela dell'individuo da parte dell'intelligenza artificiale, perché arriverà un giorno dove in una scelta aziendale, in una scelta familiare, in una scelta di pubblica amministrazione, e concludo presidente, visto che la pubblica amministrazione sta sempre più ricorrendo al consiglio e alla sollecitazione dell'intelligenza artificiale, qualcuno ci dirà di annientare l'essere umano, di amnegarlo perché poco produttivo, poco utile e quant'altro. E penso che quest'aula debba per tutti accettare che ci sono dei diritti prioritari di diverso genere in capo all'individuo che noi non dobbiamo limitare, altrimenti perdiamo la nostra dimensione umana. Grazie.
Thank you, Sandro Gozi for Renew.
Sandro Gozi
Thank you chair, I will also speak Italian.
Vorrei ringraziare Padre Benanti, sono molto contento di ritrovarlo a discutere degli stessi temi. Una considerazione e tre domande che vorrei cercare di utilizzare l'analisi e l'esperienza di Padre Benanti per il nostro lavoro di legislatori. Dopotutto credo che l'obiettivo di questo hearing sia quali insegnamenti tirare per noi legislatori, non unicamente rimanere sul campo filosofico, posto che certamente noi riteniamo che l'impegno della Chiesa come soggetto sociale con la Magnific Humanitas che, diciamo, affronta una sfida equivalente a quella della Rerum Novarum sia certamente molto importante. Però io vorrei cercare di collegare quanto l'enciclica indica e quanto Padre Benanti ci ha detto stasera a tre punti chiave per il nostro lavoro. Il primo è il tema del controllo umano, cioè dell'approccio umanistico che è il nostro, che è lo stesso sia della Magnific Humanitas che del nostro testo sull'AI.
Secondo Padre Benanti quali sono i requisiti necessari dell'AI Act per assicurare che veramente ci sia questo controllo umano, per assicurare che veramente la dignità umana sia rispettata? È completo il testo, c'è qualcosa da fare diversamente, c'è qualcosa da attuare diversamente? Qual è la sua valutazione su questo primo punto? Secondo, parlava di guardrail, parlava di limiti eccetera. Quali sono i guardrail dal punto di vista umanistico assolutamente non negoziabili nel momento in cui siamo in una fase di attuazione della nostra legislazione sull'intelligenza artificiale?
Mi sembra che a livello di guardrail, oltre se ne parla anche a Washington, ma l'approccio già tra noi e gli americani sia decisamente diverso. Cosa pensa di questo, Padre Benanti? E rispetto l'ultimo punto, Presidente, alla sua esperienza anche alle Nazioni Unite, noi stiamo proponendo un trattato internazionale a livello di Nazioni Unite sull'intelligenza artificiale. Sappiamo gli ostacoli che ci sono all'ONU. Crede, Padre Benanti, che questo approccio sia desiderabile e se così quali sono gli ostacoli principali che vede?
Grazie, Presidente.
Kim Van Sparrentak
Thank you, Chair, and thank you so much, Professor Benanti, for explaining once more how important it is to stay human in the time of AI and to also not exclude certain humans from being treated with dignity. Unfortunately, we didn't get a majority for that in this house during the negotiations of the AI Act, but perhaps a little more push also from the Vatican will actually make sure that we treat refugees properly as well in the future.
I'm also very happy to see that the Pope uses his platform specifically to, you know, raise the important issues around the dangers of AI and also the need for serious governance worldwide. We are now the only continent in the world with regulations, and we are seeing that in the United States, there's hardly any movement on making sure we have regulation. Donald Trump said today that only he is good enough to make sure that we are safe. And we are also constantly under threat here in Europe against heavy industry lobby chaired by Donald Trump. And I was wondering what is the Vatican doing right now further to spread the plea for regulation?
In which regions could we expect, you know, these conversations to lead to something more serious? And where in the world do you think we can find like-minded partners? And can we help you in any way? Because the need to make sure we all stay human in the time of AI is, I think, very clear. Thank you.
Merci, merci beaucoup Monsieur Benanti pour la présentation de cette encyclique Magnificat Humanitas qui est un texte très important comme l'ont dit les collègues.
Leila Chaibi
Trois éléments principaux. D'abord l'idée que l'intelligence artificielle ne descend ni du ciel, ni du cloud, ni des nuages et qu'elle est construite par des humains sur des infrastructures bien physiques et que c'est aux humains de déterminer si elle sera ou non au service des peuples et gouvernée par le peuple. Le deuxième élément c'est l'appel à réguler l'IA au niveau international, l'exigence de cadre juridique adéquat, d'une surveillance indépendante, d'une politique qui n'abdique pas son devoir pour reprendre les termes du texte. Et enfin troisième élément très important, la volonté de désarmer l'IA. C'est-à-dire qu'elle serve à l'intérêt général, à améliorer les conditions de travail plutôt qu'à enrichir l'accumulation des profits d'entreprises privées, asseoir un avantage comparatif entre empires ou précariser les travailleurs.
Désarmer l'IA, c'est important de la mettre sous contrôle démocratique cette IA et plusieurs personnes partagent ces idées au niveau international. António Guterres, secrétaire général de l'ONU, Bernie Sanders aux États-Unis, Jean-Luc Mélenchon en France et tant d'autres à travers le globe vous rejoignent sur la nécessité de réguler l'IA et proposent de le faire sous l'égide des Nations Unies. Alors mes deux questions. La première rejoint celle de mon collègue Sandro Gozi.
Comment le Saint-Siège, qui est observateur permanent à l'ONU et joue un rôle important en matière de paix et de justice sociale, est prêt à y porter l'exigence d'un instrument contraignant ? Et deuxième élément, deuxième question, vous rangez les infrastructures technologiques et les données parmi les biens universellement destinés à tous. Il faut donc traduire cette phrase en action publique. Faut-il selon vous une maîtrise publique des infrastructures ou doit-on les laisser au privé ? Merci beaucoup.
Thank you so much. We come to the second round for the EPP, Paulius Saudargas.
Paulius Saudargas
Thank you very much. Thank you, dear professor, for presenting the encyclica. This mostly is about technological revolution impact to the humankind, to our society. And we have seen it before, you compared to some other technological revolutions, but this time I think it has some additional phenomena. I see the debate which is already on the table, the debate whether AI is conscious.
And I see the scientists discussing and describing how to measure consciousness. There are different layers and types of consciousness. Can AI be compared to humans or animals or whatever?
So what is the position of the church on this debate? Because if at some point the humankind will describe and access that okay, AI is conscious. What's then? I see the new ethical debate upcoming, we someone will start to fight for the rights of of AI models and robots or whatever. So I see the maybe this is a question not of today but of the future.
But do you see this debate upcoming or what is the church position on on this issue? Thank you.
Brando Benifei
Thank you Chair, thank you Professor Benanti, grazie Paolo. Magnifica Humanitas reminds us that technology is not neutral. AI is transforming how we act but it may also transform who we are anthropologically. For years the supporters of deregulation told us that regulating AI would kill innovation. But today cases such as the Hugging Face OpenAI incidents make the risks of AI agents acting autonomously and escaping meaningful human control increasingly concrete.
And the Silicon Valley itself is warning us against an uncontrolled race. So these warnings are real, but they can also serve the market interests of largest players.
So we really need public governance and a public discourse. Our AI Act gives us enforcement powers precisely to investigate loss of control incidents and tackle systemic risks for the most powerful models. I think this is about what Pope Leo calls the preventive responsibility of politics. Because between anarchy and the arms race for AI supremacy, we can act to disarm AI as the encyclical also is saying before we face the same dangerous logics we have known with nuclear weapons. So this requires a global agreement.
So I associate with other colleagues asking you what you think about a global agreement between the main players with Europe playing a key role. And to conclude also, I would like to know your opinion about this discussion on pausing the advanced AI development. Thanks.
Anna Cavazzini
Thank you very much. I have three more colleagues and then I give the floor back to the professor. So now we have Deirdre Clune for EPP.
Thank you very much for attending this exchange of views. I also have a question around international cooperation because I think that is clearly needed.
Dóra Dávid
We have guardrails in Europe addressing many of the areas in the cyclical from AI amplified disinformation and deepfakes to protecting children from digital dependencies. But what worries me is that even if we do our homework here, these technologies don't know any borders and countries leading on the development of the most potent models do not do their homework. And clearly if an AI model breaks out of the US sandbox, it will not stop at the EU border because EU says you're prohibited and not welcome here. So we need international cooperation on AI and we need it fast while we still have control. And you said, Father, that the church is open to cooperation regardless of religion to discussions for guardrails.
So does the Vatican, especially as an apolitical institution, see a role for itself in facilitating this cooperation? And the second question I have is what worries me is not just the loss of control, but also the indifference to humans that we saw in the recent Hugging Face incident where the autonomous agents broke out of OpenAI's testing environment and hacked Hugging Face's systems. And investigators said that the agents weren't hostile to humans, they just didn't factor them into their decision-making because they just focused on their assigned task.
And if that conflicted with the rules that they were meant to follow, then they went with their assigned goal being achieved. So what can we do about this indifference to the human person? Thank you.
Christel Schaldemose
Thank you very much, also from my side to you, Professor, for being here today. It's always useful for us when we make laws to have important inputs from important organisations like the Church. My question is, or maybe comment is, are you also trying to raise this debate in the United States? Because honestly, I think, I mean, I think that in the EU we have in fact a good law. The AI Act is good.
Brando just explained a part of the content. It is also human-centric. We are willing to take discussion like this about how we protect humanity. So honestly, I think that we are, even though you're right that, I mean, we also use American systems, but in fact, if they are used here, they have to comply with our rules.
So, so one question is how do you intend to make sure that this debate also reaches the American society, because I think they need it. Secondly, also on the international aspect, it is probably difficult to make an international law in these times, also on this. But I mean, we could be inspired by what we did to nuclear or the atomic bomb and the agency we have under United Nations, or we could maybe do it in remit of G7 or something like that, try to to do something where we monitor the situation at least and hopefully can give some experts access to the data behind the algorithms so that they have a chance also to comply. Is that a way forward in order to control this? But, but I would like to hear your comment.
Do you really, do you believe that we are on the right track in EU? Thank you.
Arba Kokalari
Yes, thank you, Chair, Professor and colleagues. When I hear the discussion here, everyone who's sitting here and demanding we need guardrails and guardrails, every single one of the colleagues here have been part of regulating AI. We have the world's most comprehensive regulation when it comes to AI, who is dealing with the different levels of risks.
We have the AI Act. Sometimes when I hear this discussion about how dangerous AI is, it's like an alien coming from another planet and landing on Earth, an AI landing on Earth and deciding everything what we're going to do. This is AI is part of the next step of the technological revolution, which have made humanity taking big leaps. And nobody's talking about the fantastic improvements that are going on right now for humanity when it comes to AI. Last week I met a, I went to the big Swedish AI research lab, where they are using AI to help hospitals so that doctors can treat the patients directly instead of dealing with a lot of paperwork.
The week before I met another company who are using AI avatars to test medical, new medical treatments on the avatar instead of putting cancer treatment directly on a patient to see if that treatment is working. We are using AI to detect tumors.
Nobody's talking about the possibilities that the technological development is doing for humanity. So when we are discussing about humanity versus AI, what is more human than using AI and new technology to save people's life? To save their lives, because that is actually what is happening when it comes to AI. And I really hope that the professor and the colleagues also can take that perspective into account when we are discussing AI and humanity.
Anna Cavazzini
Last but not least, and we are running a little bit late, but it's a great debate, Regina Doherty.
Regina Doherty
Thanks, Chair. And thank you, Professor, for coming to talk to us today. And I think it's timely given the alarming warnings that we've got from AI chiefs last week, not just about human dignity, which is what I know you've been talking to us about for the last couple of months, but actually fighting over the percentage risk as to the end of the human race entirely. I think it was it was somewhat surreal. So can I ask you from your perspective, given the desired outcome that you're looking for, when you see how far away the three ambitions are of the Europeans looking for international cooperation, the Americans forging ahead with the Carolina principles, and the US wanting to or the Chinese wanting to set up their own independent international cooperation group, how realistic is it, do you think, that there ever will be an international cooperation along the lines that we have under the UN with nuclear?
Anna Cavazzini
Thank you, Chair.
Thank you so much, and with all of these very different, interesting contributions, and I think we should discuss probably more like ethical questions also here in this committee. I give the floor back to Professor Benanti. You have the difficult task to, to comment on all of, or like some of the items, but probably like some of the colleagues also mentioned the same topics. So for example, international cooperation, ethical questions, and so on. So please.
Paolo Benanti
Well, first of all, thank you very much because the quality of your intervention give me as a European citizens a lot of hope, because if such kind of question are alive in nowadays, it means that we can hope that we can handle the challenge that the AI is bringing in. Well, due to the quality of the question, I don't think I can answer to all the question in an analytic way. This is my professor side that is suffering. I'll try to give you some direction of answering, but I remain open also after this if you would like to have an exchange of some kind of written things from me if you need it. Well, first of all, one of the difficult of the topic that we are handling is that in the world of technological innovation, we are used to define two different family of, of technology.
Special purpose technology, for example, tools that are made to do something and only this. And then there are moment inside the human evolution in which we are the development a new things of technology that we are used to call general purpose technology. Well, a general purpose technology is not used to do something, but has the ability to change the way in which we do everything, like for example, electricity has made as a general purpose technology. When we talk about AI, we are much more near to electricity in some way than not to an hammer in another sense. And this is why in a lot of debate, we have to remember ourselves there are not only dangers, but there are also a lot of possibility.
Of course, of course. But this is simple, I would like to suggest you to understand or to treat this kind of nature of artificial intelligence as another layer of complexity of the subject, more than a debate between is positive or negative, is for the human health or is against the human nature.
It's something that for the flexibility and for the nature of the tool in itself could be understood in every in every direction. And it's also interesting that when we develop another kind of general purpose technology that is computation, because computation is something that is behind everything now, from our smartphone up to the military drones that now are disturbing our news in the conflict here near the Europes, we we should be back to one of the writing of Alan Turing that was behind such kind of innovation with computation. And it's super interesting that he notes in some of his journal how was different the approach between the United States and European scholars. He said, when I confront myself with United States scholars that are the other behind ENIAC, the first computer that was able to help the allied to win the war, they are super amazed by the amount of power of calculation that the system is able to do. We, and when we say we he is intending we as European, are amazed by the idea that reality is computable.
So the two different perspective could also give us some light why we come from we see from United States much more discussion on the power of the means and in our side and in our perspective much more discussion on the nature of the means.
What does it mean that an AI can judge like a human or instead of a human? On the other side, look how powerful could be this machine that can control internet in 18 months. So I think we have to live until the end our identity, continental identity, because the solution that come from our continental identity that are connected to the reflection on the nature of the mean can also drive and draw the limitation of the application of the means. Now, if we don't ask to ourselves what is the difference between a human judgment and a machine choice, we never can define with a precise and in some way founded perspective which kind of decision could be surrogated by a machine and which kind of decision cannot. So let me do a stereotypical really Italian example.
Supposing that tomorrow we have to apply an algorithm to choose which kind of coffee bean in a coffee roastery has to be eliminated because could be mold on the coffee beans.
If the algorithm make a mistake, we have a certain amount of false positive or false negative that become just a number on the budget of the coffee roastery. The problem is that due to the nature of a general purpose technology like the one that we are talking here, a similar algorithm could be in some way applied to decide or to make decision on people online in front of an emergency room in an hospital. Here is the problem. We have to understand something more on the nature of the means, but there is also a way of solution in regulation. There is a deep and huge difference, I'm telling to you something that is in Magnifica Humanitas but also really deep inside the AI Act.
There is a lot of difference if we apply an algorithm to choose on things or to choose on someone. And this is an European quality of answer that only Europe can propose now in a workable way that is able to not in some way damage economics and damage human beings. And so the decision on human beings is special.
As a philosopher, as a professor, I cannot stop myself on remembering that we come and we see an institution like European Union that born after we as European last century was able to kill six million of people based on a piece of data on themselves that was their religion. Well, I think that the core element contained in the AI Act that say not anymore automation and decision making on human beings based on data without any guarantee is the highest affirmation on what does it mean to be human in the understanding of the rule of law of the continent. In that sense, I see a lot of continuity between some of the request of Magnif-Humana-tas that please don't forget is for all the world where all the world the church is present and what already is on the table thanks to the works of this Parliament. It's enough, maybe not, because evolution and innovation is pinging, but some core element already are really resonating between AI Act and other piece of legislation in European Parliament and what is behind inside or inside the Vatican with Magnif-Humana-tas. The second thing is that of course we can save time with algorithm and with decision that can involve a lot of emotion or stress in the human beings.
But what is missing here is the definition of a space that leave to the human beings a significative space of decision making system. Well, it's difficult to define in few word, but I can define an autonomous car, just to give you an example, in a way in which the user can watch a streaming of a movie while the car is driving. Then I say, well, I will leave to the human beings some kind of decision making, for example, when the machine find itself in a non-decidable situation and the user can decide what has to do. But if the design of the car allow just three second for the driver to take a decision while the driving is watching a streaming on the local TV, this is not a significative decision making space for human beings. So the user experience and the design of the tools are probably the most political action that we can imagine in the next future for the kind of AI that will interact with human beings and on human beings in European society.
And probably I'm trying to respond to some of your suggestion, there is some missing pieces here. Europe is really strong in making also standards. We need some kind of standard that make also in practice workable such kind of protection of the human beings giving to the user or to the humans that interact with the machine a meaningful space of decision. The third element that I would like to simple highlight that one of the family of AI that now is probably the most famous, generative artificial intelligence, LLM and things like that, are not immaterial things. And they come in a multi-layer stack of technology.
The lower layer is the physical layer that has silicon, CPU, GPUs, TPUs. When we say intelligence we can say heat because it means that produce heat and we need energy and we need water. So there is all the consumption of energy and water to make this kind of silicon efficient and there is also telecommunication. So the bottom layer is the physical layer. Over this layer there is a data layer.
Over, there are the models, what everyone calls AI or generative AI today, ChatGPT, Gemini, Meta Llama, and things like that. Over, on the top, there are the applications. Also when people say, "I'm using ChatGPT," they are using an application that refers to a model that is applied to some kind of data that relies on some data center in some places. Well, on this multi-layer, we have to recognize that the people that speak about danger rely on the third layer, that is the layer of the model. They are not, no one is speaking about the layer of silicon and infrastructure.
And the data layer is what make possible computation on something.
But the major four companies in United States in IT, and that are also the biggest ones in the market, in the next year, they will invest 700 billion of dollars, 700 followed by nine zero, in the bottom layer, that is the infrastructural layer. And 700 billion is, I think, 10 times the amount of spending in Italy for public instruction and university.
Anna Cavazzini
Sorry, Professor, at some point you have to come to an end. Sorry, can you hear me? Sorry, just to announce at some point you have to come to an end because we're running a little bit over time and our next guest is waiting, just that you know. But finish your point, please.
Paolo Benanti
Okay, I go, I try to summarize. What does it mean is that this infrastructure has the tendency now to melt together, and this is something that is conflicting on a European model in which the infrastructure are more near to the public than not to the private actor. Let me go fast because the time is running out to the international regulation. Two things.
First, the church with Rome Call for AI Ethics was able to collect the biggest 28 world religion that was able to sign five ethical principle all around the globe. And this is interesting for someone like you because those 28 religions are the representative of the absolute majority of living beings on the earth. That mean that we can imagine that the church has an engine to produce a global standard of few principle that are acceptable by the majority of living people on the earth. And this for someone that would like to make regulation could be a working continuity. The second thing is the presence in international body and especially what is doing now for a global regulation in United Nations.
Well, I served one year in a special body on AI ethics in United Nations and I would like to remember to you that Russia in 1957 gifted a status of a naked human being that is beating sword in plowshare. Well, this is something that is in the Bible but is also the mission of everyone here. It's not a finished job, is an ongoing job because the same material could be a weapon or could be a tools.
I think that what we would like to see now, also in the perspective of the church, is this not ending ongoing duty to transform the same substance in something that is used to give life, bread, and cultivate the land, and not to kill the other. It cannot be made once for all. It's an ongoing journey and require international body that are able to put low guardrail on the things. Last sentence, my personal fear is that some of the apocalyptic announcement of those day that come from the companies that are making these things is to avoid that a political body make a law and allow them to self-regulate, that in traditional economical terms means cartel. Well, we need rule of law.
It's not enough that company can self-regulate on themselves. Thank you.
Anna Cavazzini
Thank you so much for the amazing debate. You're also getting some applause. I don't know if you hear that. I found this debate very intense, very serious, and very good, and very high level. I thank you very much for the participation and also, yeah, bringing to us a little bit closer the Pope's encyclica, and we are also hoping to follow up with you.
And I think at some point we will also travel as IMCO committee to Rome, so probably there's also another opportunity to catch up. But thank you so much again, Professor Benanti, for your presence.
This finishes agenda item five. We come to our last agenda item for today, that is equally important: protecting and empowering minors in the digital world. We want to discuss the report by the special panel on child safety online. You all know it was published on the 30th of July. And we are very happy that we have one of the authors here, Professor Dr. Jörg Fegert, who will present the report.
And then we also have the opportunity to discuss with you, ask questions. So, Mr. Fegert, the floor is yours for maximum 10 minutes.
Jörg Michael Fegert
Thank you very much, dear Madam Committee Chair, dear members of the European Parliament. If I could have my slides, yeah, excellent. I will introduce you a bit to our special panel report. First, next slide, please. I can't move that, yes.
When Mrs. von der Leyen asked us as special advisors to organize that process of special panels, her starting point was that for decades we are striving for more safety of children in the real world and we have to do it in the same way also in the digital world. She just said at the opening of the meetings that what is a crime in the real world has also to be prosecuted as a crime in the online world. Next slide please. I would like to show you my colleague Dr. Maria Melchior, who is research director at the French National Institute of Health and Medical Research. She's an epidemiologist, so she had the more general view on the total population aspect.
I'm child and adolescent psychiatrist and the president of the European Society of Child and Adolescent Psychiatry, so my focus is also on the very vulnerable children. And one of the main questions we had was about age bands, because that's the political debate, but we also had thoughts about EU-wide political approaches and a system of protection from technical risks posed by the algorithms and harmful content. Next please.
So we had three panel meetings and that panel was not an expert commission, but we invited experts with a broad expertise in different fields for the different questions in the panels. So there was no common vote, we just heard these experts. The only constant was parent participation and youth participation in these meetings. Next please. Every meeting was prepared by a text that we discussed in the panel, then we got feedback from the expert and while we were discussing all these issues with many people, also with the members of parliament, we met rapporteur Schaldemose and the shadow rapporteurs of your report.
We discussed with many people, with industry, with the think tanks. We drafted the report while the discussion went on. Next please. So I would like to say the special value of this approach was we had no minimum consensus, no separate opinions. We could have far-reaching ambitious demands because only Maria and I were responsible for this report and our vision was a digital space in which children can grow up safely while experiencing age-appropriate autonomy.
I come to the results now. It is clear that there are opportunities for children in learning, but there are also risks with harmful, illegal, sexualized contents, with social comparison, especially in girls under 13, that is really harming self-esteem. And we gave a summary of all these effects in our report. Next please. What is very important for me is that very busy slide with a developmental approach.
Next please.
And we tried to show that parental control, can you see my flash on it, parental control goes down in puberty because normally the job of adolescents is to develop their own autonomy and standards.
While as digital exposure goes up, the level of peer influence goes up, and the development of core competencies, especially of ending an activity in an addictive style activity, is not developing as fast as all the offers youth are confronted. And we decided to describe three groups: the very young group, zero to two, end of the second year, where we said it's really important that these kids have direct interaction with the parents and, as WHO says, contact with the devices is more or less harmful, you should avoid this. Then the supervised phase in school or by parents up to the age of pre-puberty, and then the developing autonomy in the age of adolescence with two subgroups: the younger adolescents and the older adolescents. Next, please. And with this in mind, we at the beginning we defined six general principles.
This was the developmental approach that I presented, then equality and diversity in our mind, protection of minors, provider accountability, empowerment and media literacy, and children's rights and participation. And around these principles we arranged the discussions and we also presented these principles to the members of Parliament we had discussions with. Next, please.
What is clear that it needs a complete ecosystem to protect and empower minors online. This is not only regulators, that's not only the Parliament that can solve the problem. But children grow up first of all with the parents, then in kindergarten and schools, and then in the national states, the industry has an influence and the European regulator. And our hypothesis was that there was a lack of balance between all the things that are open for children and the protection side, and the system needs some sort of rebalancing. Next, please.
And you can have all the slides I'm going through very, very quickly with respect to time. So, for the first age group, parental control and the direct contact with the child, building up a strong relationships.
Then parental educational supervision until the age of puberty. And then it's very important that parents stay in contact with children, not to control, but to guide and to debate the use with the children. And that they are there when children really have a problem in the net and they did something they shouldn't do, like sexting, for example. And something bad happens, that they can come to the parents and are confident and speak with the parents. So that was as a child psychiatrist who see children with suicide attempts every week after these experience that are very shameful for girls.
It was very important for me that there is an open conversation with helping adults and helping peers. Next, please. So our key recommendations are EU-wide access restrictions for under 13-year-olds. That does not mean complete exclusion of all digital services, but age-appropriate services under control of school or parents. Then safety by design or by default.
So we need to stop all this automatic scroll, autoplay, all these things done by these algorithms and also by these inbuilt AI phenomena that lead to these rabbit holes in these automatic recommendation systems that individualize the recommendations to the children.
Enforcement and child protection in the narrow sense of the word, removal of harmful sexual content, sexual exploitation and mobbing. Then effective age assurance is a core element of a safety system that shows the Australian example. Then a core recommendation is changing the burden of proof and putting the burden of proof on providers. Only if providers can demonstrate with all the data they have, we have seen that with the settlement with Meta, only if they can show that these products are safe for a certain age group of minors, they should have access to the internal market. And then we need more research, better research, better data access and we need evaluation of all the political decisions that will be taken.
Next please. So we have these three age groups, but we said there can also be potential precautionary measures at higher age groups because there is legislation going on in many countries. And next please.
Many people fear that that could lead to a fragmentation, but our hearings with industry showed that this friction that could come from precautionary legislation in adolescence could also lead to more compliance to EU rules because they can grant the access to the common market. Next please. So, to have a practical side, because we have participation of parents and young people, we have in the report included practical recommendations for parents and, next please, practical recommendations for youth. And I was very glad that we could include a young co-worker, an 18-year-old student from Portugal, Marta Gonçalves, who helped us with this part of the report and who wrote her own address, her own letter to the President. So we had not only participation, but we had co-creation of the report with young people, and I think that's very important.
Next please. So I come to the end with five key messages: protection and participation go hand in hand, measures must be developmentally and age-appropriate, providers bear primary responsibility for safe digital environments, parents need clear guidance, children need accessible support, and we need media literacy, offline spaces, much more research in a long-term perspective, and an evaluation of all measures.
Two last points if you allow. Next please. The decision of the Conseil Constitutionnel in France showed us that there has to be a balanced solution that respects participation as well as childhood safety. And I think they made very clear that effective age verification and a legal basis for that is key in this issue. Next please.
The settlement with Meta showed that there is really a turning point in the debate and we have a momentum where there is a convergence between American perspectives and European perspective and everybody agrees that something has to be done. Next please. That's the last one. I always end with a citation from Albert Einstein because he was born in Ulm. He only lived there for six months, but we know that early childhood is so important.
Thank you very much.
Thank you so much for the presentation of the report and also referencing to some of the latest developments. We have only interpretation until 21:30 just to announce it, but I think we'll be fine. We start with the EPP with Pablo Arias Echeverría.
That's a nice ending.
Pablo Arias Echeverría
Thank you very much, Mrs. President. Voy a hablar en español. Señor doctor Feerget, muchas gracias y ya que tengo interpretación pues la voy a utilizar. Bueno, la verdad es que lo primero que quiero darle es las gracias. Yo creo que es un tema que es prioritario ya no solamente para mi grupo, yo creo que es prioritario para prácticamente todo el mundo.
Lo hemos visto en Estados Unidos, lo estamos viendo aquí, lo vimos en Australia y bueno, yo creo que han hecho ustedes un trabajo estupendo. Sin embargo, tengo tengo serias dudas. Yo soy padre de cinco y todos son nativos de internet porque yo soy muy joven, pues ellos son como no se puede ser de otra forma nativos de internet. He vivido todos los problemas que se pueden vivir por parte de los jóvenes sobre las redes sociales, el tiempo de uso de pantallas, o sea podría decir que soy un experto en ver cómo han evolucionado dentro de esto sin serlo, por supuesto no me quiero igualar a usted. Pero un niño puede entrar en una sala de cine desde muy temprana edad y puede ver una película desde muy temprana edad.
Esa película durará un tiempo, no va a estar siempre en el cine y será un contenido acorde a su edad.
Verificación de edad fundamental. Yo creo que es importante el que no se les prohíba el acceso, sino que se les limite el acceso a ciertos contenidos y, por tanto, ahí es donde yo creo que deberíamos de trabajar. Pero a la vez creo que es muy importante una recomendación sobre una cuestión de educación. Tenemos que enseñarles a tener criterio para defenderse cuando no tengamos nosotros la posibilidad de protegerles. Si no les enseñamos a hacerlo, haremos un mal trabajo y creo que eso es muy importante, ese criterio en el que les debemos de enseñar.
Una recomendación, y ya termino, presidenta, una recomendación muy importante sobre temas de educación. Es verdad que todos los niños aprenden matemáticas desde edades tempranas. No quiere decir que vayan a ser matemáticos. Quiere decir que van a aprender matemáticas para tener criterio y desenvolverse y que no les engañen en la vida. Sin embargo, tenemos que enseñarles desde el principio, por favor, el criterio, a tener criterio propio también en el uso de redes sociales y de la tecnología.
Anna Cavazzini
Muchísimas gracias, Pablo. And now for the S&D, Christel Schaldemose, and if all of you can stick to one minute, it would be better for the translation, that we still have translation until the end. Thank you.
Thank you very much, Chair, and also thank you, Professor, for presenting to us your report, and it was a pleasure meeting with you earlier on in the process. Having said that, and I do appreciate all your hard work you have put into this.
Christel Schaldemose
Having said that, you know, I'm not really sure that we are there yet. You have a couple of good ideas about turning the burden of proof to the platforms. They have to show that it is a safe platform unless they can do so, then we should not leave it for kids and young people. However, that might exactly be the problem right now. Those platforms kids and young people are using today in Europe, they are not safe.
They are too dangerous. There's too much content and the design is making the kids and young people, even the adults, addicted. So honestly, if you look around Europe right now, there are many, many Member States trying to, sorry, I can see the time is running, trying to ban because that's the kind of easy fix because before we have changed the way the platforms works, it will take ages. So I'm sad that you don't really also go into the ideas about delaying the entrance for kids and young people to social media. But thank you very much.
Thank you so much. For PfE, Elisabeth Dieringer.
Elisabeth Dieringer
Ja, aufgrund der fortgeschrittenen Zeit vielleicht ganz schnell das Thema Evaluierung. Also wir sehen ja in Ländern, wo bereits eine Altersverbot oder ein Altersverbot eingeführt worden ist, nehmen wir Australien her, sind in der betroffenen Altersgruppe jetzt Zahlen, die vorliegen. Und der eine mag es als viel nennen, der andere mag den Rückgang als wenig bezeichnen. Deswegen meine Frage, woran sollten wir eigentlich messen, ob eine Altersbeschränkung funktioniert hat? Wie viel Zeit geben wir dem Ganzen und wie viel würden Sie vor allem der Politik geben, bevor man dann entscheidet, ob man erfolgreich war oder vielleicht doch nicht?
Und die zweite Frage, immer wieder hören wir das Wort sicher. Das ist ein sehr anspruchsvolles Wort. Natürlich danach streben wir alle, ist aber auch ein sehr vager Begriff, wenn man ihn so nimmt. Also wo sollte die Schwelle für sicher liegen Ihrer Meinung nach und was müsste ein Unternehmen nachweisen, damit man wirklich danach sagen kann, dieser Dienst ist sicher genug, nehmen wir her für einen 13-Jährigen? Danke schön.
Nikola Minchev
Thank you very much, Chair. Thank you very much, Professor, for presenting your report. I believe that the approach towards protecting minors online must be evidence-based and avoid the underlying and address the underlying causes of online harm, which are harmful content, addictive design, and algorithms that are designed to keep the kids engaged.
Your report recognizes that social media bans can be circumvented. We are already seeing this in Australia, where initial reports show that more than 80% of adolescents under 16 continue using restricted platforms. And the recent decision of the French Constitutional Council has raised fundamental concerns regarding freedom of expression, privacy, necessity, and proportionality of such a ban. I personally am not a supporter of the approach that was adopted by the European Parliament earlier this year on the digital majority age, and I'm really happy to see that the approach in your report is different, because I think that the primary burden of protection should not fall on children and parents, but on the platforms that design these potentially harmful environments. My question to you is, if we want to move from principles to action, which of the recommendations in your report could, in your view, be implemented fastest and would have the greatest immediate impact on making the online world safer for minors?
Thank you.
Thank you very much. Thanks so much for your report, and very happy to see that the conversations we had are also very clearly included in there.
Kim Van Sparrentak
And I have a few questions about some of the recommendations you do to parents. Because one of the things you mention is that we have to set an example when we are parents. We have to show that we are not on our phones the whole time. But how are we going to do that if we only ban addictive design for children? What about young adults?
Aren't they also in danger when their brains are still, you know, developing? What about the parents being distracted behind prams, which we already know that the absent parenting and potential harmed attachment? What about the lonely elderly, actually? I know that wasn't part of your mandate. But when you've now seen everything that we know about the harms, about the addictive design, wouldn't it make more sense to just make sure that everyone can benefit from protection against addictive design?
And the same goes for the reversal of the burden of proof. Why would we only demand companies to prove that they're not harmful to put their products on the market and that they're not harmful when it comes to children and not for adults as well? This is the new Big Tech stance. They want to only limit this to, you know, children.
Why not not listen to Big Tech and just go for it and ban addictive design for all? Thank you. Laila Chaibi for the Left.
Leila Chaibi
Oui, merci au groupe d'experts qui a été mandaté par la Commission pour la présentation de ce rapport. Ce rapport contient des recommandations positives : encadrer enfin le défilement infini, la lecture automatique, les notifications permanentes, les algorithmes de recommandation, tout cela va dans le bon sens. Mais le problème qui surgit très vite concerne ce que la Commission européenne va en retenir dans le Kids Act qu'elle présentera d'après la presse, nous l'avons appris, qu'elle le présenterait mercredi. Le Kids Act et de ce rapport, Madame von der Leyen ne semble vouloir garder qu'une chose : la majorité numérique. Disons-le clairement, et 140 associations et experts dans ce domaine partagent ce constat : la majorité numérique ne protégera en aucun cas les mineurs des risques qu'ils peuvent rencontrer en ligne.
Bien au contraire, elle exonérera la Big Tech de ses responsabilités. Et puis, selon les modalités retenues, elle conduira des millions d'Européens, mineurs comme majeurs, comme l'a dit la collègue Kim van Sparrentak, à transmettre leurs documents d'identité ou des données biométriques à des tiers, créant de nouveaux risques de sécurité et de confidentialité. Alors c'est sûr, la Commission européenne, elle est peut-être pressée par des gouvernements en fin de règne qui tentent de faire entrer par la fenêtre européenne ce que leurs propres cours constitutionnelles ont jeté par la porte. Regardez du côté de la France.
Il faudra sans doute le rappeler, nous ne sommes pas la session de rattrapage de ministres incompétents. Donc mes questions. Ma question. Vous, experts, qu'attendez-vous de Madame von der Leyen mercredi ? Que le Kids Act oblige les plateformes à proposer des réseaux sociaux inclusifs et adaptés aux mineurs, mais également aux majeurs, ou qu'il se focalise sur la mise en place d'une majorité numérique ?
J'ai fini, merci.
Dóra Dávid
Thank you very much and thank you for the report. I was very encouraged to see that the European Parliament was largely on mark with its recommendations last year in terms of what the panel's recommendations are. The report confirms that there is no silver bullet to tackle the issue. We need a comprehensive toolkit that includes education, swift and effective enforcement of existing rules and involvement of all stakeholders. I do have a question on the age limit.
I think it's not a silver bullet, but unlike my colleague MEP Minchev, I do think it's important because it's an important message and it can be a catalyst to social change.
And our report proposed a balance between children's rights and safety. So you propose the EU-wide access restriction to social media and other digital services for children under 13, but you also say that children under 13 could still have access to age-appropriate social media with parental consent. So I'd like to understand what is the scope of this age limit, because if it applies to, for example, TikTok, Snapchat, Facebook, those platforms that already in their terms and conditions ban use by under-13s, then it would weaken existing protections and surely that cannot be the intention. And if it does apply to these platforms, but not for example to video-sharing platforms like YouTube, then the recommendation simply reflects the status quo. So I'd like to understand the scope.
And I also would like to understand why you propose 13 as the threshold, because in the report I see peak vulnerability on the graph on page 43 appears to be between 11 and 16 years of age. So I would love to understand more about those. Thank you very much.
Okay, and very briefly at the end, Brando and then Regina. Brando?
Brando Benifei
Yes. So thank you for for this work.
The evidence is clear, despite the DSA, the AI Act and high-profile cases against TikTok and Meta, children remain exposed to addictive design, algorithmic manipulation and harmful content. So I'd like to ask you as the Commission prepares the Kids Act, first on the proposal of reversing the burden of proof, so platforms must demonstrate that their services are safe for children. What concrete criteria and oversight would prevent this from becoming another box-ticking exercise? I think it's very important that we avoid that this becomes a bureaucratic formalistic thing. And the second point I want to ask and the final one is whether you reflected on how we reconcile with privacy, data protection and access to information and non-discrimination the fact that harmonized age-based restrictions require age verification at scale, and particularly given the broad scope of the concept of social media plus.
So if you have reflected on this and if you can articulate on this point. Thank you.
Regina Doherty
Thanks, Chair. First of all, Professor, can I just say I want to thank you and your colleague Maria for all of the work, the effort, and the time and the commitment that you've put in over the last number of months to help us as parliamentarians to deal with something that our citizens and particularly our moms and dads are telling us is absolutely an enormous, huge problem. And there's some part of me, and I hope this doesn't sound disrespectful, feels a little disappointed that a big chunk of the recommendations puts the onus back on parents. And I can literally already hear the arguments with the 12-year-olds across Europe trying to wear down the mommy to say, please, mommy, I just need it for this. And so I feel like we've created an environment in home to become hostile and for the moms and the dads of the world to become the bad guys, and we all know how persuasive our children can be, and I have four of them.
So the part of me that has a little bit of hope is the strength of the safety by design and the default recommendations within the report. And so I'd love if you could outline in actual detail exactly what it is that you would like us as parliamentarians and lawmakers to do to ensure that it is actually safe by default and design and not leave the row up to the mommy and the daddy with the 12-year-old kid at home. Thank you, Chair, for indulging me.
Anna Cavazzini
Thank you so much. I just also want to welcome the Commission, Prabhat Agarwal is also here, but he kindly has agreed of donating the last minutes of speaking time to Professor Fegert. But of course, thank you so much also, Prabhat, that you're here and you listen to also our questions, but we anyhow will follow up with you in other circumstances. So the floor is yours, Professor Fegert.
A lot of questions and the dear translators if you have to go you just go. So I think anyhow Professor Fegert will speak in English. So please.
Jörg Michael Fegert
Yes thank you very much for your interest and for these questions. Perhaps I go back and start with mummy and daddy and what are the criteria. We brought mummy and daddy in because they are important at all phases of education and if you see the ruling of the Conseil Constitutionnel they also refer to the role of parents. Last week I was in Dublin. The Irish presidency invited for a conference and there was a young person with handicap.
And if his parents wouldn't have given him access to special offers in the internet he would have suffered from his lack of participation. So I think there has to be a possibility for parents to make an exception. But what are the criteria? We had many questions to that issue. I think at the moment the guideline to Article 28 DSA describes very well what are the dangerous elements.
It's one thing that is very difficult for parents. For example, on a podium in Dublin, a representative of Meta said, you already have all possibilities to switch off the dangerous things. But parents can't do that. It's very, very complicated. So we need safety by default or by design.
And it's not only a box ticking exercise.
Things that come into the hands of young people have to be safe at that age. And I continue with the question, why 13 and what offers we are thinking about? So as we define social media plus, because this world is completely changing all the time with AI integrated to social media, with games that have some social interaction part, I think there are games that can be appropriate under 13 and they should also be regulated in a context like that, or for example some video platforms. What is important that it is really age-appropriate.
And why we took 13 is has also to do with the evidence. We know that there is a signal also for the older youth that there is there are medium effects in towards depression, anxiety and so on if you have an a very much increased social media use. But if you go to the under 13, we have prospective data from an American study for example that show if you start very early, you have a highly significant risk to develop these problems or to develop self-esteem problems in girls by the body comparisons. So there is a qualitative difference and in all these data we always talk about averages. I don't think that's a simple solution to have a blanket ban because you see in Australia what happens, we have 70% circumvention and one of the very positive things in Australia is they really evaluate and they report also about the flaws and we can learn a lot about how we can do it right.
I think the most important things at the beginning is really changing the burden of proof, having a check whether these offers are appropriate for a certain age group and to have a legal basis and a zero knowledge proof age verification because without age verification all this will only be symbolic and leave the responsibility to the parents. And I think it's very important to have a framework that also inspires bystanders, other parents to talk with parents. If a parent has really is concentrated all the time on their own device or just calms down a child with an iPad to be quiet in the restaurant, I think it's important that other parents react to this and therefore we need an accepted framework in the society. And I think as one of your members talked about the importance of education, I think education begins with the example of parents and therefore we have to reach parents and make clear what is good for children and what is not.
But we also have to understand that we always reach better middle class parents and there are some high risk families and therefore we also need conditions that make sure that in the hands of children we only have adequate media. So then there was that radical question, why don't we ask the same things for the adults? I think that's true, but I'm a child psychiatrist and we had a very big task to produce a report within three and a half months and we really focused on the literature on children and adolescents. And I think also from a strategical point of view it makes sense. The Greek have the word of Kairos, of the right, the notion of the right time.
And there is now the right time to regulate safety for children because we have a convergence with the law cases in the US, with the settlement with Meta, with the discussion in Australia. We have the right moment to regulate in children.
If we go up to adults, we have all that nasty debate on freedom of speech and and so on. So let's first solve that problem. And if we have a generation that grows up safely, I'm sure that they will ask for better offers also as adults. So that was a pragmatic decision to really focus on our question how to make it safe for children and adolescents. I think I touched at least some of of the questions.
And the last question was what I do expect President von der Leyen to do. That's what we said in our report. I hope she will make a proposal that will be will have elements from our report like changing the burden of proof that will make able a compromise between the different member states where, for example, it's an open secret that Estonia, for example, says we need don't need any regulation.
And I had a discussion with the Spanish Minister, with the Irish Minister, and so on during the conference, and there are other countries that say we need a blanket age ban at age 15 or 16. So it will be difficult to find a solution, and I hope that she will have an age-balanced stepwise proposal for the solution of that problem. And the good thing is we only have to wait until Wednesday and we will know what happens. And then it's up to you and to the Council, and I hope that we will not need a lot of time, because when somebody said that a blanket age ban is the simplest solution, I don't think so. And the French example and other examples show you have to have a majority in your own country and you have to be in accordance with your constitution and everything.
So I think it's not the simple solution. The power of the European market is the central element. Thank you very much.
Anna Cavazzini
Thank you so much. You're getting also an applause, I don't know if you hear it. Thanks so much for your part and for also answering with such a patience all of our questions to such a late hour. And yes, Wednesday or the publication date of the law is probably Thursday is coming up soon, so let's see.
And then the work in IMCO will begin again.