Committee on the Internal Market and Consumer Protection Ordinary meeting
Legenda
Marcadores da transcrição
- texto Frase em reprodução
- texto Resultado da pesquisa
- texto Intervenção citada pela ligação que abriu
Clique numa frase para reproduzir o vídeo a partir desse momento. Selecione texto para o citar com o orador, a marca temporal e a ligação.
Mais opções
Desfasamento
Desfasamento da transcrição
Exportar
Captado em direto
Versões deste vídeo
- Gravação Versão definitiva · 14 de julho de 2026
- Captado em direto Transmitido em direto a 14 de julho de 2026Em exibição
As transcrições podem diferir ligeiramente de uma versão para outra.
IMCO European Parliament
Seguir as suas sessõesAnna Cavazzini
Who's against? Abstentions? Adopted. We go to compromise five. Who's in favor?
Who's against? Abstentions? Adopted. We go to the final vote by roll call. I open the vote.
Your vote, it just, your vote doesn't work? Oh shit. One second, so let's not close the vote. Could you check this from a technical perspective?
Is it working now? Can we repeat the vote or is it not possible? So sorry, we break off the vote and restart the vote again. Probably then it will be working. So we close the vote, I open the vote again.
Okay, Katrin, we add your vote in the minutes as an abstention, okay?
Anna Cavazzini
Okay, so we add it in the minutes that you abstain, but it's of course not good that your voting machine is not working.
Let's hopefully try that you can solve it for the next vote. Ah yes, sorry, sorry, and the result. 34 in favor, 13 abstentions, and 3 minus. Congratulations. I'm so sorry, I was abstract.
Okay, now we come to the next vote on the draft opinion on the proposal for a regulation suspending the application of the rules on the appointment of an authorized representative for extended producer responsibility for batteries and waste batteries and packaging and packaging waste. Our rapporteur for the opinion is Klara Dostálová and she's not here, so I will not give her the floor and we would basically directly proceed to the vote. So we start with draft opinion PA and 1, they're identical, so the first vote will be 1. By roll call vote, I open the vote. I close the vote.
Yeah, so this is not adopted. That means we proceed with the vote and start with compromise 1. Who's in favor? No, 1. Compromise 1.
Who's in favor? Who's against? Abstentions. Rejected. That means we now vote on compromise 1A.
Who's in favor?
Who's against? Abstentions? This is adopted very clearly. That means we go to compromise 2A. Who is in favor?
Who's against? Abstentions? Let's check that. I open the vote. This is rejected with 29 against.
One second. So now we go to 2B. Who's in favor? Who's against? Abstentions?
Let's check. I open the vote. I close the vote. This is adopted. That means we go to the final vote.
So there was a lot of like if that go that, but now we go to the final vote by roll call. I open the vote. One person's missing. Everyone has to vote.
I close the vote. That means it is not adopted. Sorry, there was some internal discussion. I tried to get the information, but it's rejected. So we don't have an opinion.
That means we go to the next vote. That is item six, the general budget of the European Union for the financial year 2025, all sections, where we vote on the draft opinion first and then the vote on the budgetary amendments. Our rapporteur for the opinion is Piotr Müller, and I first give the floor to the rapporteur.
Piotr Müller
Ja może tylko krótko powiem, że w tym sprawozdaniu przede wszystkim chciałem podziękować też za współpracę między wszystkimi kontrsprawozdawcami. Natomiast pokrótce zależało mi przede wszystkim na tym, aby w tej komisji wybrzmiały kwestie związane z ochroną naszego rynku, wzmocnieniem instytucji, które są związane z kontrolą celną, z tymi wszystkimi instytucjami, które zabiegają o to, aby faktycznie nasze wspólne zasady rynku wewnętrznego były przestrzegane. Jest wiele innych jeszcze oczywiście elementów tego sprawozdania, ale to myślę, że każdy je doskonale zna. Dziękuję bardzo.
Dziękuję bardzo.
Anna Cavazzini
Sorry, thank you so much. So as I said, we start with the draft opinion. And again by show of hands, compromise one, who's in favor? Who's against? Abstentions?
We go to compromise three, who's in favor? Who's against? Abstentions? Adopted. We go to the final vote by roll call vote, I open the vote.
I close the vote. 41 in favor, 5 against, 4 abstentions, so congratulations, it's adopted. And now we go to the budgetary amendments for the 27 draft budget, and we vote all the budgetary amendments one by one by show of hands. We start with number one, who's in favor?
Who's against? Abstentions? Adopted. We go to number two. Who's in favor?
Who's against? Abstentions? Adopted. Number three. Who's in favor?
Who's against? Abstentions? Adopted. Number four. Who's in favor?
Who's against? Abstentions? Adopted. Number five. Who's in favor?
Who's against? Abstentions? Adopted. Number six. Who's in favor?
Against? Abstentions? Adopted. Number seven. In favor?
Against? Abstentions? Adopted. Number eight. Who's in favor?
Who's against? Abstentions? Adopted. Number nine. Who's in favor?
Who's against? Abstentions? Probably let's check that because that was not so clear. I open the vote. I close the vote.
Okay, adopted. Very well. Number 10. Who's in favor? Who's against?
Abstentions? I think there was a clear majority, so adopted. Thank you so much. That concludes the budget votes.
And now we have still some voting on the provisional agreements resulting from interinstitutional negotiations from the Omnibus IV, so two files. I give the floor to the rapporteur first, to Reinier van Lanschot.
Reinier Van Lanschot
Thanks, chair. This is a joyful moment because this Omnibus is one of the Omnibuses that knows only winners. 400 million per year will be saved for companies, is the expectation of the European Commission, by stopping the obligatory paper manuals to be delivered with household appliances and other products. So that's a massive cost reduction, but it's also better for consumers because they will have more access to digital manuals and the security instructions will still be physically present. On top of that, it will save a huge amount of trees and as somebody who loves nature, that's something that makes me happy too.
So, let's all vote in favor and I want to thank everybody again for contributing to this Omnibus because it's one of the best ones we've done in the past year, according to me. Thank you.
Thank you very much. As you have seen, we have two votes because it's technically two files. We start with the digitalization and common specifications, so the agenda item seven.
Anna Cavazzini
By roll-call vote, I open the vote. I close the vote. 45 in favor, four against, zero abstentions, so congratulations. But we're not done yet. We have to have a final vote on the, it's also called digitalization and common specifications, whatever, on the agenda item eight.
I open the vote. I close the vote. And with the same majority adopted, so congratulations. This concludes our voting session, so we do one minute technical break and then we continue with our exchange of views with Anthropic.
Okay dear colleagues, let's restart. Sorry but I wanted to have everyone out of the room who is not taking part at our IMCO session. But now I think it's good if we sit down again and move our bilaterals to outside if you want because we still have some very important topics in our agenda before hopefully all of us can have some rest in the summer holidays. We had been as you know inviting Entropic several times to the committee and I'm very happy that today we are welcoming Mr. Donny Greenberg, member of the Entropic technical team via video I think.
So, welcome digitally. Yeah, because Anthropic has been in the headlines for a couple of months because of different reasons that you all know. And we thought in the IMCO committee it's very, very important to also have an exchange of views on that, on the implications for Europe, on the implications on cybersecurity, but also the question of how much we are in the European Union dependent on other countries' presidents' decisions over technology. So I'm very happy that we can have this exchange of views today, and again, welcome Mr. Greenberg. So at the beginning, we will hear from you for around seven minutes, and then of course the members can come in and ask questions and make comments.
You can already signal to the secretariat if you want to speak. But first I give Mr. Greenberg the floor to present the company, but also to present how you see the present situation. Thank you.
Donny Greenberg
Chair and honorable members, thank you for the invitation to speak with you here today. Allow me to introduce the company I work for. Anthropic is an AI safety company founded in 2021. We're headquartered in San Francisco and build a family of AI models called Claude. We're a public benefit corporation, which means we're legally bound to weigh the public interests alongside the interests of our shareholders.
Our purpose is to ensure that the world makes a transition through transformative AI safely. This is why we publish our research on where these systems are becoming dangerous and why we hold ourselves to a responsible scaling policy that ties deployment to demonstrated safety. We were among the first signatories of the general purpose AI code of practice and have championed the adoption of regulation at both state and federal level in the United States.
Before I dive into the details of today's discussion, I'd like to point out that I'm a technical person, not a policy person. My background is AI training and serving systems and since I've joined Anthropic, I've co-led Project Glasswing to help the cybersecurity world navigate the emergence of AI offensive threats. To find myself speaking with you today is both an unexpected honor and also a clear indication of the importance that AI and cybersecurity play in the safety and security of our citizens. Cybersecurity is a clear example of the dual-use nature of AI capabilities. We did not train Mythos Preview for cybersecurity but rather for coding and general reasoning.
But a model that writes and understands code well is inherently good at finding flaws in software systems. In testing it found thousands of previously unknown vulnerabilities including in every major operating system and web browser. It found a 27-year-old flaw in OpenBSD, one of the most security-hardened operating systems in the world, as well as several bugs in the Linux kernel, which it was able to chain together to gain control of the machine. We've shared more detail about the critical vulnerabilities Mythos Preview has found in our red team blog. We concluded that releasing the model broadly would be irresponsible and so instead we began Project Glasswing, a company-level effort in partnership with cyber defenders to prepare for the emergence of AI cyber offense.
We gave model access grants and support to the organizations that build and maintain the critical foundations of much of the world's software systems that much of the world's software systems depend on, so that defenders could find vulnerabilities, shore up their defenses, and innovate to discover new ways to use the models for defensive cybersecurity. Last month, we extended the program to roughly 150 further organizations in more than 50 countries. Within the first month, our partners had found more than 10,000 high or critical severity vulnerabilities. Mozilla found and fixed 271 in a single Firefox release, 10 times what they found in the previous one. Finding vulnerabilities is no longer the bottleneck.
Fixing them and incorporating AI capabilities across defensive security programs is. Triage and remediation has become a major bottleneck, and other areas like incident response, penetration testing, configuration scanning, and patching all urgently require attention and creativity to give cyber defenders a durable asymmetric advantage with AI. Within three to six months, we expect many other AI companies will have models of this class, and some may release them without safeguards. The window in which defenders hold the advantage is measured in months. Finally, I wish to address the export control directly.
On the 12th, the US government applied export controls to our newest models, Fable 5 and Mythos 5, requiring us to restrict access by foreign nationals.
Anna Cavazzini
Excuse me, Mr. Silverback, could you talk a little slower because the translators have difficulties in following?
Donny Greenberg
Oh, I'm sorry. Thank you. Requiring us to restrict access by foreign nationals. We suspended access for every user.
The controls were lifted on 30th of June and access was restored on the 1st of July. The past month has made us all acutely aware that the technology we're creating is bringing with it additional geopolitical complexity. We welcome our cooperation with the United States government in order to identify a sustainable framework that offers our customers and governments around the world predictability. We also welcome our cooperation and constructive conversations with the AI Office, the European Commission, and ENISA during this first-of-its-kind moment in AI. Let me end with what I find encouraging.
The same capability that makes these models dangerous makes them the most powerful defensive tool that has ever existed for this problem. If wielded properly, we feel strongly that this technology favors defenders long-term. If I had one message I would want to be remembered from our discussion today, it's that now is a moment for organizations to strengthen their cyber resilience, drawing from both well-established best practices to novel experimentation and industry collaboration. We have watched closely as the EU has announced its AI cybersecurity action plan, and as other governments around the world also hold discussions on this important topic. I expect a difficult period ahead.
But on the far side of it, we see a world where critical systems are significantly hardened to offensive threats, and cyber defenders maintain a durable advantage over bad actors.
No one organization can reach that goal alone. Coordinated, highly collaborative action across many organizations will be necessary. Frontier AI developers, software companies, security researchers, open source maintainers, and governments are all essential to navigating this moment to secure our digital future. We're encouraged to see the spirit of mission-driven collaboration already playing out across the cybersecurity field. Again, I am grateful for this opportunity to discuss this important topic with you today and I look forward to your questions.
Anna Cavazzini
Thank you so much. So, as I said, we have now the opportunity for the members to come in and ask questions and also give some remarks for roughly two minutes each and we start with EPP with Dirk Gotink.
Dirk Gotink
Thank you very much, Chair. Thank you, Mr. Greenberg, and for your very quick introductions.
I think, I mean, we've talked a lot here about AI in the EU and speaking EU, I don't know where you are, maybe in San Francisco, but if you're speaking from there, the world looks a lot brighter from the AI perspective than if you look towards San Francisco from here. Because basically your very noble mission of accompanying Europe or the world in the AI transition has become a geopolitical tool and an economic tool by the intervention of the White House. This has of course shocked and actually emphasized once again that maybe that partnership we have from a transatlantic point of view is not as solid as we always expected it, but on a technological front, it is actually quite obviously now one of the US leading AI and the EU trying to catch up. Now I have first a general question. Would Anthropic be able to be developed in the EU?
Some people argue that in fact LLMs and all kinds of other AI models are not able to be developed in the EU because of a regulatory environment that does not allow for it and also some energy-related issues. Would you have a reflection on that? Secondly, did anything change in what you, what was blocked as a product at the first intervention of the White House and then what was freed again in June? What has changed in the meantime in terms of the actual program itself?
Was there was there something on the inside that was did you make changes to it or not? And of course you can tell us whatever you want. We don't really have the the means here to check it, but it would be important for us to understand what is the what has changed in the meantime apart from from that decision. And secondly, how do you foresee in the future to to roll out newer models without this kind of circus where we first need a president stopping it and the rest of the world then being very, very concerned that in fact we're only giving advantages and security advantages to American companies as opposed to trying to protect everyone at the same time. Thank you.
Christel Schaldemose
Thank you very much. Also thank you for being here with us online today. Well, maybe along the same lines as Mr. Gutink, I'm I'm wondering if you would sit in in our chairs here in the European Parliament, wouldn't you then consider I mean whether it makes sense for us to use Anthropic Anthropic's business model because at one hand it creates kind of a dependency on you and at the same time we would also generate a lot of growth and a lot of data for Anthropic.
So wouldn't you recommend us that it's better for EU to try to develop our own systems rather than using yours in order to have strategic independence, but also in order to keep the economic value here in EU? If you were here with us, would that be your consideration? Thank you.
Kim Van Sparrentak
Thank you very much. Interesting to hear your perspective. I guess you can expect a lot of policy-related questions that perhaps you won't be able to answer as the technical guy, but we're nonetheless going to ask them. And I think Anthropic could have known that. So we heard about Mythos that it's basically able to detect all cybersecurity vulnerabilities.
And if that's true, then it could pose a major risk, especially if it falls into the wrong hands. But even if half of what is claimed is true, it opens the door to major cyberattacks at scale that can really, you know, harm our banks, our hospitals, and other transport systems.
And we know that Mythos is just the first model. And the biggest worry I guess we have in Europe is that we have a US government that doesn't believe in regulating AI, and therefore we are in the hands of Anthropic to decide whether they think the EU will be a trustworthy partner or not, and also to decide who they will give access to and who is worthy of protecting against the dangerous model you have built. So, and now we've also seen that the US has of course weaponized this model against the rest of the world, or has tried to do so. So what is Anthropic exactly doing to ensure people and businesses are safe, not only in the US, which I know that you do care about, but also in Europe, and that the model doesn't fall into the wrong hands, and that this model won't be weaponized for geopolitical gain? I think it's really important for the European people to know that you are not developing something, which you don't have to, you can also stop developing this model, but that when you are developing a model, you're doing that in a way that is actually responsible and safe and won't be weaponized against the European people.
Thank you.
I take one more and then probably do a short answer round and then we continue with the round of questions or remarks. I had the next one on my list Laila Chaibi.
Thank you.
Leila Chaibi
Thank you. I'm going to speak in French so maybe you can press translation. Thank you. Oui merci beaucoup monsieur d'avoir accepté de de venir devant notre commission. C'est quand même dommage que vous soyez pas là physiquement mais bon merci d'accepter l'échange en ligne.
Alors effectivement nous on a été stupéfaits ici en Europe suite à la décision de Donald Trump d'empêcher tout citoyen non-états-unien d'accéder à vos modèles Phable et Mythos. Et tout ça ça nous a prouvé quelque chose d'essentiel c'est que pour garantir notre cybersécurité l'Union européenne on ne pouvait ici se reposer sur des modèles étrangers américains ou chinois et vivre sous la menace permanente d'un kill switch à la guise du gouvernement américain états-unien mais ça aurait pu être chinois par ailleurs. Alors une fois cela dit j'ai un peu la même question que certains collègues c'est quel a été le deal ? Maintenant nous avons à nouveau accès à Phable vous l'avez dit et donc en échange de quoi avez-vous permis l'accès à Phable ?
Est-ce que vous pouvez nous dire les modifications que vous avez réalisées sur votre modèle qui ont permis que le gouvernement des États-Unis d'Amérique autorise son exportation ? Ma deuxième question, c'est pour me faire l'écho de plusieurs articles qui sont parus dans la presse et qui nous disent que votre entreprise aurait espionné via des outils de tracking des utilisateurs de Claude Code qu'elle considérait comme chinois pour empêcher que ses modèles soient copiés, et cela sans le consentement des utilisateurs ciblés. Alors ma deuxième question est celle-ci : est-ce qu'Anthropic, sous pression ou non du gouvernement américain, procède à de tels trackings à l'encontre de citoyens européens ou de citoyens qui pourraient avoir la double nationalité tout en vivant aux États-Unis ? Merci.
Anna Cavazzini
Merci beaucoup. So now Mr. Greenberg, I give you the floor back because otherwise we're piling up too many questions. So please, the floor is yours.
Donny Greenberg
Um, so I, uh, I'll just run through these questions in order. Um, so yes, I'm, I, uh, appreciate you having me, regret that I couldn't be there in person. I'm joining from New York. Um, to the, to the first question about could Anthropic be built in Europe, um, I, uh, to be honest, I don't really have a strong answer to this question.
Anthropic is a very particular company and culture, and I don't think that the topic that we're talking about today is really specific to Anthropic or me in particular. The moment of preparing for cyber offense, I think is independent of individual companies or people. I think that the message that we are trying to send with Project Glasswing is that AI capabilities are coming to cyber defense and cyber offense imminently across the globe. And we expect our competitors to have models of similar capabilities quite soon. And so I don't think that the question should be about whether building Anthropic in Europe, whether Anthropic could, you know, this specific company could be built in Europe.
It's really about what we can do globally, especially as a cybersecurity field, to collectively prepare and collaborate for these capabilities to arrive across a number of providers, some of whom won't be providing open weights or won't be providing them with adequate safeguards.
To the second question about what is changed following the export controls, over the course of the period the export control period, we collaborated closely with partners inside the United States government to review and test our safeguards on the model and to assure many stakeholders, public and private, of the robustness of our safeguards and then even increased the the recalls we say, the robustness of the safeguards as an added precaution. But to be clear, that was only on our Fable class of models. The Mythos class of models, which are the ones of note for dual-use cyber capabilities, were not the subject of that work as they have limited safeguards. There wasn't a specific change before and after to the Mythos class models. To the third question about how we are evolving rollouts in the future to avoid this, we are absolutely working tirelessly to avoid these types of complex staged rollouts in the future.
To be honest, I think when we initially began when when we were faced with this dilemma of how we deliver the capabilities of Mithos Preview to cyber to cyber defenders um without having to bear the risks of widespread cyber offense with the model, it was very clear that it was going to be a long road for us to build up the the access programs and protections to be able to roll out the model as broadly as we would have liked initially. And so, you know, it was a combination of actually needing to build out those protections and programs in addition to actually being prudent so that we could analyze the risk as we progressively rolled out access, just recognizing that there were unknown unknowns and that the um these novel capabilities might be the subject of interest by cyber adversaries and that we would need to monitor closely to ensure that they weren't being misused. We are building up to be able to roll out models in the future in a way that we can deliver the capabilities to cyber defenders precisely and in a targeted way and with a with a clear understanding of the risks so that we don't have to do this type of staged rollout in the future.
And also please please stop me if I if you want to call time or move on to other questions. To the question of whether it is better for the EU to develop their own systems, I think that our message is to my initial point that really not about Anthropic's model specifically or to you know any one provider or any one geography's models within this moment, it's really that these models generalize and that over time the models will continue to grow in capability. And it is not our objective for it to be the case that the the technology delivers economic benefits only to the people who hold the model but rather that the technology delivers economic benefits to everyone. And so from a cybersecurity standpoint, we think it's critical for cyber defenders to be evaluating their security programs from the lens of a diverse set of models and the possibilities of cyber offense in light of that, especially ones that might be open weights or don't have adequate safeguards. And and I don't think that the the moment is is one of specifically trying to build you know one one special or magic model that is going to provide a special you know distinct type of cyber offense, it's really about upleveling cyber defense globally in a collaborative and creative way to give asymmetric advantage to cyber defenders.
And we spread that message to everybody. You know, for many companies who have access to Mythos Preview, they use a pretty diverse set of models to actually do cyber defense. For example, many companies use generally available models like Opus 4.8 to do triage and patching as opposed to using Mythos Preview because they find that it's highly efficient for those use cases. And so in general, we advocate for governments and companies to be building up these capabilities broadly across their security programs, not to over-index on a specific model or provider. To the fifth question, what we are doing to ensure that people and businesses are safe globally, this is central to our objective for the Glasswing program.
The way to think about the methodology of the program is that it is about balancing risk.
The risk of granting a new organization access to the model, including the risks of compromise or misuse that go along with that, against the risks of not granting access to the model to that organization and not giving them as much uplift as we could to their cyber capabilities that we could to shore up their defenses or to improve their cybersecurity programs. And this is not a view with a lens on a particular country or region. It really is a global view. The software foundations that we target or that are the highest priority, at least for vulnerability discovery, are ones that underpin the entire global economy and that a compromise would not discriminate to a particular region. And our safeguards and the systems that we design to avoid or detect misuse or compromise are also designed to protect everyone equally.
They're designed to ensure that access remains only with critical cyber defenders and that the model can't fall into the wrong hands.
We have highly elaborate and complex defense-in-depth systems to ensure that the model access is protected at many levels and that the critical work of the infrastructure maintainers, the maintainers of financial infrastructure, physical infrastructure, communications, digital, etc., that they can focus on up-leveling cyber defense to protect us all.
Anna Cavazzini
Probably now we can do a next round of questions, or you're still having a lot of things to say in this round?
Anna Cavazzini
Sure. Okay, great, because then I can take in probably more colleagues. Thank you so much for your answers so far. The next one on my list is Pablo Arias Echeverría for the EPP.
Pablo Arias Echeverría
Thank you very much, Chair. Voy a hablar en español. Mire, lo primero, darle las gracias al señor Greenberg por estar aquí, pero tengo una duda y es la primera pregunta que le quiero hacer. Y es muy fácil. ¿Está usted respondiendo a las preguntas que le estamos haciendo o está simplemente leyendo lo que nos responde su modelo de IA?
Me ha parecido que estaba usted leyendo, por tanto, me gustaría que me respondiera. Yo no he venido aquí a atacarle, he venido a tener un debate, pero sí que me ha sorprendido esto. La inteligencia artificial yo creo que no es mala por sí sola, lo que es mala es la utilización de la misma y, por tanto, creo que puede haber propósitos espurios o peligrosos por gente que tiene mucha intención de hacer caja o de tener poder a través de la IA. Mire, la pregunta es también bastante fácil, aunque creo que no es fácil la respuesta.
¿Hacia dónde podemos ir para que la tecnología sea por el bien de la humanidad? Esto ya lo hemos leído y lo hemos visto, por ejemplo, cuando decían los fundadores de OpenAI al principio de esta iniciativa que era por el bien de la humanidad y ahora estamos viendo que no es precisamente ese el fin en el que están ahora mismo. Y no solo para el beneficio de unos pocos, que es lo que parece, ¿no? O algo que debería ser positivo para la humanidad, como digo, como por ejemplo en el ámbito de la medicina, pues no lo sea solo para el ámbito de la defensa, en este caso para el ámbito de la guerra. O sea, esto ya lo hemos oído antes.
Por el bien de la humanidad nos lo decían en la era digital y estamos cometiendo los mismos errores que antes. Estamos esperando a que esto empiece a coger velocidad de crucero y lo que nos estamos dando cuenta es de que los gigantes de internet son los que se benefician de esto y, además, esto ya es indispensable para todos. Entonces, nosotros también hemos vivido antes ejemplos que creo que serían fundamentales y aquí está mi pregunta. Hemos tenido acuerdos internacionales, por ejemplo, con las reglas de seguridad vial. Y ha ido muy bien.
Una luz verde, todo el mundo pasa, una luz roja, todo el mundo se para. Una señal de stop es full stop y una señal triangular es precaución, igual que circular es prohibición.
No se puede buscar hacer algo parecido con la IA antes de seguir avanzando en algo que puede convertirse en extremadamente peligroso para la humanidad? Esa es una pregunta que creo que es muy fácil de difícil de responder, pero que creo que tiene sentido el hacerla y creo que es fácil formularla. Muchas gracias.
Thank you so much. So the next one for S&D is Ms. Guzenina.
Maria Guzenina
Thank you, dear Chair. I'm also IMCO rapporteur on CSA2 and NIS2. So it's really nice to hear the answers Mr. Greenberg is giving us here today. The cybersecurity map in Europe is very fragmented. And this is one of the problems because when you have one weak link, it allows the penetrators to infiltrate our system.
And I would like to hear your opinion, Mr. Greenberg, on how would you strengthen cybersecurity as a whole from the European perspective? Obviously, we have different member countries. Some have already taken steps. We have different kind of threats. I come from Finland.
We are constantly under hybrid attacks from Russia, for example.
So we need strengthening already ASAP. And then there are countries that are kind of still dormant, maybe not taking seriously enough the threats we're having. And then we have threats to the banking system, for example. And they are developing their own cybersecurity. And at the same time, there is talk about Q-Day.
So question number one: how would you strengthen European cybersecurity as a whole? And then, is Q-Day ever going to arrive, or is it going to be a race, like in arms, that we're just going to develop better and better defense systems so that the criminals who are at the moment trying to get hold of the information that they cannot at this moment encrypt, and we're just going to build a stronger defense and at the end of the day, they will never get to us? What is your opinion?
Thank you. Now I have Virginie Joron.
Virginie Joron
Donc aujourd'hui nous interrogeons Anthropic, fleuron américain de l'intelligence artificielle, et ce faisant nous touchons du doigt, non, nous heurtons de plein fouet le grand paradoxe européen. D'un côté des modèles comme Mistral d'une puissance si per- capable inédite débarquent sur notre continent. De l'autre, quel est notre réflexe en Europe? Le bâton. Le bâton du flicage par le bureau de l'IA, le bâton des amendes qui tombent dès le 2 août, les bâtons de nouveaux plans directeurs de la Commission pour verrouiller l'accès à ces technologies.
Virginie Joron
Monsieur le représentant d'Anthropic, en tant que députée européenne du groupe Les Patriotes pour l'Europe, je défends moi avant tout la souveraineté des nations européennes face à la bureaucratie de Bruxelles et à la domination technologique étrangère. L'AI Act voulu par la Commission impose des règles lourdes et idéologiques qui risquent d'étouffer l'innovation, on l'a vu, européenne, tout en laissant des géants américains comme vous prospérer sur notre marché. Comment jugez-vous cette approche réglementaire? Ne pensez-vous pas qu'elle handicape dangereusement nos entreprises et nos États au lieu de protéger finalement nos citoyens? Nous voulons une Europe des nations souveraines capable de se doter d'une véritable autonomie technologique.
Êtes-vous prêt à nouer des partenariats directs avec les États membres sans passer par les instances bruxelloises pour développer une IA qui respecte notre identité, nos valeurs et nos intérêts stratégiques ? Concrètement, seriez-vous disposé à localiser des données et les traitements en Europe, à garantir une totale transparence sur vos entraînements et à adapter vos modèles pour qu'ils servent d'abord la sécurité et la prospérité de nos peuples plutôt qu'à la mondialisation sans frontières ? L'Europe pour vous ne doit plus être un terrain de jeu pour les Big Tech américaines. Quelle est votre position sur une véritable souveraineté numérique puisque c'est un gros sujet ici à Bruxelles ? Je vous remercie.
Kamila Gasiuk-Pihowicz
Thank you. Following the recent decisions of the United States administration to restrict access to the most advanced AI models for non-United States citizens, Europe is asking itself a fundamental question. Can we build our security, our competitiveness, our digital future on technologies that may become unavailable because of political decisions taken outside Europe? So let me start with this. How much can Europe count on Anthropic as a long-term partner, not only for innovation, but also for cybersecurity and defense?
And how can we be sure that access to your most advanced technologies will never become a geopolitical bargaining chip? Secondly, how can you guarantee that European institutions, businesses, and citizens will not suddenly lose access to key technologies because of decisions taken in Washington rather than in Brussels? This leads to a broader question. Who ultimately decides where the limits of your models lie: Anthropic, democratically elected governments, or the US national security apparatus? Where does corporate responsibility end and government influence begin?
Finally, I would like to address the issue of data protection, because trust is impossible without transparency. Can European users' data be accessed by US authorities? Please give me a clear answer: yes or no. Where is European customer data stored, and who, both technically and legally, can access it? And finally, if United States authorities requested access to a European customer's data, ordered restrictions on your services, or required changes to the way your models operate, would you inform that customer?
Thank you.
Reinier Van Lanschot
Thank you, Chair.
Thank you, Mr. Greenberg, for taking the time to talk with us. I noticed in your introductory remarks that you are not really specialized in policy but more in the technical details, so I'll frame my questions a bit more to the technical side, although I must admit that I also find it a bit disappointing that in this exchange we cannot touch upon that policy side, because that is actually where I think a large part of the essence of what we need to address together lies. You mentioned that your purpose as Anthropic is to make a world transformed to AI safely, and I also notice that in your statements of the values that you find important, you try to be an AI company that also has a certain degree of ethics. But to what extent do you think it's possible to uphold that when autocrats and autocratic regimes could hijack your mission and use it for their benefit?
Is there anything you would change in the structure of Anthropic to make sure that that risk is decreased? I'm also curious about energy and water usage. As we know, these data centers use a lot of energy and water, and I think one of the important things for the future of AI is to make sure that the water and energy usage decreases. So, as a technical expert, do you think there can be any breakthroughs expected that diminish the amount of water and electricity that is used? In Europe, but also in the US, the energy grid is overstretched already, and we don't want this to go at the expense of new houses and other companies that need energy and water as well.
And last but not least, how important is it for the strategy of Anthropic to have more of their compute power in Europe itself and not in the US alone? Thank you.
Thank you. Now I have Brando Benifei.
Brando Benifei
Thank you, Chair. Mr. Greenberg, thank you for this opportunity. In April, Anthropic gave 50 organizations early access to a model that finds vulnerabilities at an unprecedented scale, but not one was European. In June, the US administration suspended the access that had just been offered to ENISA. It was then restored in July after Brussels lobbied on Washington.
That sequence tells Europe what it needs to know. We need to have more digital sovereignty, and we are working on that also with proposals on the table like the CADA.
But we also need to look more in depth. The tool our defenders depend on is granted at your discretion and withdrawn at someone else's. This is what happened. MITOS has surfaced more than 10,000 critical vulnerabilities. Anthropic says that the bottleneck is now patching.
European operators of power grids, hospitals and water systems run the same code as our partners and they will learn of the flaws only last. That is a European security problem created by a private access list. So my question is very simple. How many European entities hold access today and on what conditions? Is ENISA's access working now and who can take it away?
And this: you opened your models to the British Institute, which cannot regulate you. But from the 2nd of August, the AI Office can, with the competences it was granted by the AI Act. Will you offer it and ENISA the same access before your next release? Europe should not trade enforcement for access. We will apply our rules and we want you to comply.
This occasion is very important to have a frank dialogue. I want to thank you for that, but we need answers on all these points.
Anna Cavazzini
Thank you very much.
Thank you, and the last speaker from the MEP side is José Cepeda, also for S&D.
José Cepeda
Sí, voy a hablar en español. Bueno, yo, en primer lugar, agradecerle que esté aquí presente porque la verdad que hace ya tiempo que estamos buscando algún tipo de explicación a los modelos nuevos de Anthropic, de Mytho, todo el ámbito de la ciberseguridad y la ciberdefensa con estos nuevos modelos está, está sinceramente quebrado si no hay una respuesta alternativa. Hemos visto cómo OpenAI ha creado su propio modelo. Mi pregunta es si usted tiene algún tipo de conocimiento que en el ámbito chino DeepSeek está ya trabajando sobre modelos parecidos o semejantes, ¿no?
Y luego respecto a Anthropic en concreto, me gustaría saber qué límites desde el punto de vista ético y operativo y militar sucedieron en la operación que el ejército norteamericano desarrolló en Venezuela para que Anthropic fuera apartada de todos los diferentes sistemas de inteligencia artificial militar de Palantir sobre aquella operación y posteriores, ¿no? Yo trabajo en la Comisión de Seguridad y Defensa y estamos muy preocupados realmente de que este tipo de modelos de desarrollos de LLM de carácter ofensivo, en este caso de carácter militar, utilizando ya el nuevo modelo de inteligencia artificial agéntica, es decir, sin control humano, pueden implementar sistemas militares de alto riesgo sin ningún tipo de control de control humano, ¿no? Y en ese sentido me gustaría por último preguntarle, en línea con lo que algunos de mis colegas han puesto encima de la mesa, qué garantías jurídicas vinculantes puede dar Anthropic a la capacidad europea en un contexto de que no pudiera ser revocable por una decisión unilateral de un gobierno tercero, en este caso, por ejemplo, el modelo norteamericano, porque el modelo que han puesto ustedes en Claude para uso y libre disposición rápidamente ha sido, digamos, cerrado, capado a uso público por una decisión de la administración norteamericana. ¿Qué garantías tenemos en Europa si empiezan nuestros gobiernos a utilizar Anthropic de carácter defensivo de que una decisión unilateral del gobierno norteamericano no nos pueda dejar fuera del sistema? Muchas gracias.
Anna Cavazzini
Thank you so much.
So with that I give the floor back to Mr. Greenback and you heard our concerns and a lot of critical questions and probably yeah one feedback. I mean we're very, very happy that you're here with us but also a lot of MEPs would have liked to speak to the political level. Probably you can also take this back for the next exchange of views we're definitely having. So the floor is yours again.
Donny Greenberg
Yeah, thank you. And I appreciate all of the all of the thoughtful questions. To the first point about whether it's me speaking or Claude speaking, I suppose I should take that as a compliment because we're very proud of Claude but it is certainly me speaking. I'm just frantically taking notes because there are so many questions and reading from my notes. I want to I think there were a few themes across the questions that I want to just address in order.
I think that the technical capabilities of the model actually do provide quite a bit of quite a bit of input I should say to some of the policy some of the policy questions that are being asked. So I want to start with the question about how infrastructure especially you know very remote or regional infrastructure can defend itself in this moment even without direct access to the model.
And I think it's really important to understand that the model is not a magic shield nor a magic spear. It's not having access to the model does not grant an organization sudden defensive capability that they are impenetrable. The model's capabilities in cybersecurity are much more subtle than that. If you were a hospital in Finland and you had access to this model, you don't necessarily maintain first-party software to analyze with the model to look for vulnerabilities to patch and improve your cybersecurity posture. You depend on vendors for that software, and those vendors are global.
As does a hospital based in New York, the vendors that it depends on are global. And the people who have to steward that software, who are responsible for using the model to identify those vulnerabilities and to improve the security posture of hospitals and power grids everywhere, they're the ones who need to use the model to shore up those vulnerabilities. And so I think the important thing to keep in mind is that these security problems are really hard.
They would actually be dramatically simpler if we could just give people the model and then they would suddenly be defended wherever they are. The legacy software underneath a huge amount of the world's physical infrastructure is very hard to update and has many existing known vulnerabilities, let alone sophisticated vulnerabilities that advanced AI models might find. That is arguably a prerequisite or a more fundamental problem to our global security posture than whether a particular power grid uses a model directly. This isn't to say that it's not our objective to extend access to as many cyber defenders as we can, that is absolutely our objective.
The point is to say that there's a huge amount of collaborative work that globally we must take on to identify all of the vulnerabilities across this very interconnected global and digital economy that we have, to be able to use the model or whether using the models or not, identify creative ways to respond to these novel threats. And so I think that that is really the principal message, that there is a huge amount of scrutiny that we must apply throughout our software infrastructure, our physical infrastructure, to identify the path forward. There's a huge amount of collaboration that needs to happen. I'm quite encouraged by already a huge amount of sharing that happens across security organizations from staunch competitors in a field to help one another shore up the infrastructure or shore up the software that they need to help their own customers to improve their security posture. And so I think that there are a number of policy implications that flow out of this.
For one, it's more important to be highly coordinated and to be highly collaborative and to be part of this global security conversation and to identify the key vulnerabilities at both the software level and at the operational level as far as what can be patched, how quickly it can be patched, what is already, you know, who has the resources or whether the resources are present to be able to upgrade the software. Being a part of that conversation to uplevel cyber defense broadly is I believe the most crucial thing.
And we've seen in some industries a huge amount of that collaboration, which has been great, um, independent of geography. And in other industries, we haven't seen as much collaboration, or it's been harder, um, it's been harder to coordinate, or the problems are more challenging, either the software's older or there's less resources, um, or the systems are more sensitive to downtime. And so I think these are the most fundamental questions that need to be addressed first. Um, I think that it's also important to recognize that the, the, the security stance that any, um, country systems will have in the face of, um, uh, novel AI offensive threats, um, is going to really depend not on whether there's a, you know, single, uh, model in question, um, which was used to scan a particular piece of infrastructure. It's really going to be a matter of preparedness to all models and a, a matter of preparedness to the moment of, excuse me, of AI cyber offense.
And so I would say, if it's a question of are we picking the right model or do we have the right model in Europe, I actually think that looking at only a single model leaves a lot to be gained as far as the opportunities to analyze the behaviors of different models or even to take advantage of differences in capabilities across different models. I don't think that it's wise to just focus on a single model's capabilities or using it or owning it in a particular way. I think we generally advocate a pretty diverse and holistic approach to cybersecurity, especially in the face of AI threats. To the first question about whether AI will serve humanity, this is our express objective. Our mission, especially our mission with Project Glasswing, is for to do everything we can to help these models serve humanity by navigating towards a world where cyber defenders have a permanent or a highly durable asymmetric advantage.
And that can only be done in a highly collaborative, highly coordinated way, full stop. It's simply not possible for a single entity like ourselves or for a single country's cybersecurity industry to solve this problem.
It's a global problem with global interdependencies and global concern. Our objective is to allow these models to eliminate the risks of cyber attack to all critical infrastructure, to all critical, you know, financial, physical, civic, and digital infrastructure. And I think that the point about having one weak link allows penetrators in is a good one. It really needs to be, this needs to be analyzed at all levels. You can't view it as simply as, well, you know, we depend on these few software vendors or even we have, you know, we as a country have these few software vendors as our, you know, our, you know, flagship tech providers.
If those are protected, then we're protected. It's simply not true. It's just a dramatically harder problem that I know many in this room are already highly aware of and have well-coordinated industry groups and we've talked to the AI office about this before. The level at which...
Donny Greenberg
Yeah. We are eagerly working closely with those groups to analyze these threats at all levels to make sure that there isn't a single weak link.
I'll give a simple example. In many, in physical infrastructure like power grids and water systems, you have a distinction between your OT systems and your IT systems, where the OT systems control the actual functioning of the physical infrastructure in question. And the connection between the two is at the communications layer. So even if you think that your software for running your business as a power grid is secured because your vendors have assured you that they are, if the communications layer is the weak link, then that could allow offensive threats to traverse into the OT infrastructure. You have to be extremely precise at all levels to understand where these weak links are and to understand the global picture and interdependencies.
And our objective is for we as a model provider and for governments and for private organizations to collaborate closely to identify these weak links, excuse me, and prepare for a time when models of this capability are more broadly available. And so I think I want to, I'll close with that as a positive picture.
I feel that the spirit of collaboration in cybersecurity is extremely strong. I meet with a huge number of CISOs across the globe and they remark about how this moment is one of intense focus and intense collaboration to identify these vulnerabilities and to prepare for AI as a, AI coding and cyber capabilities broadly as opposed to, you know, focusing on, you know, just their one company or just their one region. It has really been a global effort which is quite encouraging. Thank you.
Anna Cavazzini
Thank you so much, thank you so much. We still want to of course also give the Commission representative the floor. I welcome Lucilla Sioli, Director from the AI Office in the room, correct? Ah yeah, there you are. Hi, good to see you.
Lucilla Sioli
Hello, thank you very much and your honorable members, thank you very much for this exchange. It was also very interesting for us obviously from the Commission to follow your questions and the replies given. I'm the Director of the AI Office and I would like to start simply by reminding that very soon on the 2nd of August the enforcement powers of the AI Office under the AI Act will apply and that will help in making sure that these models, at least when they are deployed in the European Union, are safe.
And we therefore have important safeguards and ways of checking that the mitigation measures put in place by the different companies are satisfactory from the safety point of view. But I also want to remember that now, also in the United States, there is a regulation on these models because I would like to remind that there is an executive order which invites the providers of the most advanced models to designate them and to have them checked before they are placed on the market. And so I think that this element of a first piece, if you want, of regulation on AI in the United States is important to keep in mind. In Europe, we also just launched an AI and cyber action plan on the 7th of July. And it's an action plan that on the one hand strengthens our capacity to test and evaluate these models, in particular when applied to cybersecurity context, but it also provides a blueprint to organize access to the models.
And that's going to be, I think, very important for the future and for the relationship also with the providers in terms of the choice of the trusted users. So with the action plan, we also try to ensure there is a cyber hygiene and that there is patching ongoing. We heard it doesn't have to take place through MITRE, it can take place also through other tools. And what is important is that we will act to avoid creating new dependencies and always only relying on non-EU frontier models. Today we have focused very much on cyber capabilities.
I want to remind that these models increase their capabilities and they may bring about other kinds of risk, for example in biology, and that we must prepare to be able to live and face and live safely with these capabilities. So all in all, it's not just a question of competition in tech, it's also very much a question of technological sovereignty and security. And I very much welcome the attention of the Parliament to this matter. Thank you.
Yeah, thank you so much also for being here and also answering to some of our questions and also of course giving the perspective of the AI Office on this important matter. So thanks again to our external guest from New York, Mr. Greenberg, for participating and of course also to all the members that raised questions and concerns.
Anna Cavazzini
I think it was a very important debate. This concludes agenda item nine and we come to agenda item ten, one of our hits, the common charger that we did in the last term, where Alex Saliba was the rapporteur and we now, as we announced also at the beginning of the term, want to have more scrutiny sessions to also check how the Commission is actually implementing our legislation and to see what effect our legislation has. So it's very important that we today focus on the Commission's report on the common charger directive and see what the results are and probably also what the gaps are. We first have the Commission and then we have our our rapporteur and of course then also members can come in and ask questions and take the floor. And for the Commission, I welcome Anna Xavier, head of unit from DG GROW, to present the report.
Welcome. Hi.
Okay, good afternoon everyone. I don't know if we have the presentation uploaded.
Anna Xavier
And thank you, Chair, for the intro. Perfect. So thank you very much, thank you, Chair, and thank you all for the opportunity to be here today with my colleagues Alexei and Nikos to present very briefly the report, as you say, on common charger, though it has a much bigger name, in fact on market developments, fragmentation and technological progress of wireless charging for radio equipment. The report is online and I hope you all had the chance to look at it. If we can go forward.
I don't have the... Yes, thank you. So this is a report prepared under Article 3, item 4 of the amended Radio Equipment Directive.
As you said, Madam Chair, when we were here discussing in 2022, there was a lot of interest in possible future extensions of the so-called common charger. And therefore, the co-legislators introduced an article to have a regular report that would assess the market developments, fragmentation, and technological progress with the potential to identify categories of products that would be suitable to apply the common charger requirements. These requirements, of course, the most known one is, of course, the type USB-C receptacle, but of course, there are other elements like the unbundled sale of the charger and the device, and of course, some information labels and pictograms, so some informative elements that come with the products under the scope of the amended RED. The idea is and the aim of having more products or the analysis to have more products is to see if there are products included that could improve further consumer convenience, reduce waste, and avoid market fragmentation.
As you all know, there is already a very wide range of products covered by the directive, you know, the phones, of course, and tablets that we know best, but also more recently the laptops as of 28 April this year. Therefore, if we continue with the next slide, what we did was to commission an independent support study to do the analysis along those lines that were very clearly identified in the article of the directive and to very simply try to explain here, which is not maybe such a simple analysis, but of course we start by looking at products that are within the Radio Equipment Directive scope. We look at whether they are rechargeable, that is, whether they have a battery that can be charged via wired charging. And then we look at, this is basically more technical aspects, and then we have of course the more economic aspect, which is of course whether they represent a significant and stable or growing share of the market of sales, so if their sales are growing, and also if the products that could potentially be within scope are an important share of the segment.
Of course, also we had to look at whether there were technical barriers. So whether we could, de facto, it would be possible to implement the requirements. And we looked into four main categories that we list there. These are the very four main, let's say, market segments under Radio Equipment Directive: consumer electronics, smart appliances, toys and wearables. I think a lot of the consumer electronic products are already part of, under the scope of the amended RED.
So and many of these already apply the common charger requirements. So we looked into additional categories. And if we then move on to the next slide. What we first had, we had first nine categories which fit, which would sort of be could be identified on the basis of this technical, economic and technical and economic sort of, well, criteria.
Thank you very much. But then of course we also looked at other criteria which led us to exclude two categories: the electric toothbrushes and the remote control toys. These were mostly reasons related to safety. Basically it was deemed unsafe to have the electric toothbrush under these requirements because they are used in wet environments. And on the toys because of course there are safety concerns about of course those who play with toys that are children and of course it's important to keep them safe and it was deemed that this would be better then to keep the charger coming specific charger coming with the toy.
So, all in all, what we have are these seven categories you have there, three under consumer electronics, four in wearables. We also do a cost-benefit analysis where we look at the costs for manufacturers of adapting the products to fulfill the requirements, and compare that with the consumer savings and also with the benefits, environmental benefits. What you can see is overall we have some consumer savings, which are less or more important according to the category. We also have some costs for manufacturers, again, depending on the category they are high or lower. And we also have some environmental benefits that we identified, but of course, when we compare these with the initial set of products that we analyzed under the common charger, the both the benefits and costs are much more, are much smaller.
So, they are relatively small compared to the initial set of products. So, this is where we stop. Thank you very much, and yes, I pass you the floor, Chair. Thank you.
Thank you so much for the interesting implementation report. You finished, right? Or any of the colleagues want to come in?
Perfect. Then I give the floor to Alex Agius Saliba, who, as I said, was the rapporteur last term.
Alex Agius Saliba
Thank you, Chair, and thanks, dear colleagues, and also thanks for the Commission for this important study. I am happy, first of all, to hear that the directive has made a positive difference for many ordinary citizens, and I believe that this should be a landmark of the priorities that we should maintain here in the European Parliament and also in the Commission and Council when it comes to priorities, priorities which change the lives of our consumers for the for the better. As we have read in the report of the Commission published last week, 84% of Europeans expressed a positive view of the common charger, and we have between 55 and 65% who are in favor of continuing to extend the scope.
The numbers in the report show us why. The benefits for European consumers can run into tens, tens of millions of euros if we apply the common charger standard and common charger requirements to other consumer electronics like laptops and also wearables. And this comes on top of the non-material benefits and convenience to not have to carry on multiple chargers for different devices. Despite the USB-C becoming a standard, we cannot leave it just to the goodwill of producers to be able to implement it as it will slow down the uptake and could also fragment the internal market. We have a legislative instrument in the directive and we should and we must continue to use it.
And while VR headsets, drones, video games, controllers and wearables like smartwatches are a smaller part of the market, it has been rapidly growing, making regulation relevant also from a consumer perspective when it comes to extending the scope of the common charger.
Especially as the Commission report has found that extending the scope to these categories entails limited adaptation costs for manufacturers. So two questions from my end. How will the Commission turn these crystal clear conclusions into action and when we can expect proposals to enlarge the scope of the common charger directive? And my second question, could you elaborate on what you are planning to do to enforce the existing requirements as your study found that over 60% of device owners still experience difficulties charging them with another charger. Does the Commission see this as a problem and if yes, how do you intend to rectify the situation?
Thanks a lot.
Thank you so much. Is there any other member who wants to come in? Dirk.
Dirk Gotink
Thank you very much, Chair. I think this common charger is a is again a proof that every time you do harmonize something really well in the in the in the internal market, it actually works. And the difficulty is every time overcoming the vested interests who are making a lot of money with the fragmentation and with the protectionism that exists. So this is a great example of that. And I mean I share the same questions as Mrs. Saliba.
I would have another additional question and that is more the voluntary adoption of the standard considering that it simply makes sense for for producers to to start producing the same type standard that most people have in their kitchen drawers when they need to sort of charge their devices.
Are you seeing that as well in the market? And if so, is that going in a, is that a quick adoption and would that also maybe circumvent the need to legislate or is a simply a forced legislation on all these product sectors still necessary for that adoption? Thank you.
Anna Cavazzini
Thank you so much. I put myself on the agenda. I don't see you guys, no. So yeah, like my two colleagues, I'm very happy that the first report of the Commission also shows that the law is a success and I guess everyone who also campaigned in the European elections can confirm it. It's a consumer topic that a lot of people noticed from the European Union and that doesn't always happen.
So that's very good, not only for consumer convenience but of course also for the environment. And similarly to Alex, I'm also wondering about the extension to other categories and what your plans are there. And as you probably remember, in the negotiations we also had big discussions on the obligatory unbundling to even increase the environmental impact. And I wonder if you have some data on it. So is it still that because now consumers can decide, so is it still that consumers ask a lot to have the cable directly with it or is there also any data that shows, okay, probably the unbundling, the obligatory unbundling could make sense?
And if no other member wants to come in, then I give the floor back to the Commission.
Anna Xavier
Yes, thank you very much for for your questions. So at this at this stage, we we well, actually, we put forward the report and we came here also to hear from you and to hear feedback from, of course, the various stakeholders. As said, we do find benefits, we also find some costs. They are relatively small compared to what we had originally. So perhaps it's not such a as compelling a case as it was before.
And therefore, bon, we are a bit in a hearing mode to also see a bit what what are the views of the various stakeholders before putting forward any legal act.
Of course we can. It's in the directive and we can, if we want and if there is a compelling case, then we can put forward a delegated act that changes the annex and includes these additional products. I think that one important point is exactly what was also mentioned, which is a bit the voluntary adoption of the standard. And what we see also, if any of us, for example, now goes to try to buy, for example, a smartwatch, you already see that many models, not all, but many models have adjusted to some of these requirements at least. So therefore, indeed, the question is, would this be enough and we expect them, then, manufacturers to also adopt all the requirements, or do we need to apply legislation?
So I think that is an important aspect which we are currently reflecting upon. We have also our expert group on radio equipment where we will present the detailed report and we will hear from market surveillance authorities, notified bodies and industry, so also to better understand how the market is reacting because, you know, technology is moving faster than we can produce analysis. On monitoring and enforcement, I guess this is the same as with everything else that is under the directive. So we have market surveillance and of course if we identify some issues, then, well, if market surveillance authorities identify some issues, then under the parameters of the directive then we can also take action. But before we take action, I mean, there's a lot of room for national market surveillance authorities to act.
On the unbundling, we are also conducting another study on unbundling. It's still on and I we still don't have any precise results to mention. It's still running until I think the end of the year, no?
Yeah, until the end of the year. Did I forget to say something, Alexis or Nikos? Thank you.
Anna Cavazzini
Thanks so much for the answers, but then yeah, we're looking forward to some of the studies and some of the facts you're still looking at and of course to potentially some of the delegated acts. This concludes agenda item 10 and we come to our last agenda item for today, the Circular Economy Act, driving internal market harmonization, boosting sustainable business models and enhancing EU resilience. We put it on the agenda because the act will be soon published and we wanted to check if the Commission has already some preliminary indications of where the travel is going, especially of course also regarding the the harmonization and single market parts that will be definitely in. So I welcome very much the Commission that will present where we stand and then afterwards we have the opportunity for a discussion or for questions.
So I welcome Stefano... no, not yet here. Wait one second. So I was told now the Commission is arriving, that's great. So we can get first indications of where the Circular Economy Act is going and probably we can also get an update on the calendar so that we are best informed as committee and then we can also ask questions. And I welcome very much the Commission representative Stefano Soro, Head of Unit for Sustainable Products and Textiles from DG GROW, and you have seven minutes.
Welcome.
Stefano Soro
Okay, thank you very much. Thank you, Madam Chair. Thanks to the IMCO Committee and the members who have given us another proof of the non-partisan, across-groups, across-committees interest in this initiative, the Circular Economy Act. The circular economy for the European Union chimes with one Europe, one market, less waste, waste more value. It's an imperative from many points of views.
Climate and environmental is the most intuitive, economic and financial. We are a small, densely populated continent with not so many natural resources and with an increasing, facing an increasing pressure on the raw material markets. It's an imperative from the point of view of political and geopolitical stability and sustainability, a crucial lever of strategic autonomy. And last but not least, this is a dimension which is often overlooked, but I think it's very important, it's an imperative from the point of view of social sustainability because it's linked to high value-added employment which is harder to delocalize. The political context of the Circular Economy Act is in the political guidelines of President von der Leyen, the mission letters of EVP Séjourné and Commissioner Ostwall and the Clean Industrial Deal.
There is a clear interest reflected in these policy actions by the President and the Commission in the circular economy for the reasons that I mentioned a second ago.
The main goals that we are setting ourselves with the Circular Economy Act are, on the one hand, simplification. We want to unlock the scale of the single market for circular products and materials and beyond. So we are considering measures on the harmonization, simplification, and digitalization of EPR schemes. It's sometimes disheartening to see that it's easier to bring a product on the internal market from wherever it comes with one representative anywhere on the internal market responsible for its safety, its compliance, its certification, etc., etc. But often, if your product is subject to an EPR scheme, you need an authorized representative in every Member State, and you can see very easily what sort of impact this has, in particular on small and medium enterprises.
We also want to simplify the shipment of recycled materials and of critical streams of waste or critical raw material-containing waste by harmonizing, simplifying, especially harmonizing and introducing mutual recognition clauses for end-of-waste criteria.
The second high-level goal that we have is competitiveness. There is an emergence of critical raw material recycling value chains in Europe which need more feedstock and increased predictability on feedstock availability. It's before dealing with the Circular Economy Act, because of my experience, I was not familiar with waste legislation and I discovered to my puzzlement that waste legislation mentions exports and cross-border operations referring to the internal market. The result of this and of a lack of harmonization is the fact that it's often more expedient and easier and more remunerative to get rid of valuable materials outside of the European Union rather than keeping them on the internal market. So we have more, we want to make more feedstock available for other recycling streams.
We want to create or facilitate rather supportive conditions to the emergence and development of new sectors, in particular refurbishment, which has a positive impact on job. Repair, reuse, refurbishment are mostly local jobs that are more not impossible, but more difficult to delocalize. The treatment of critical waste streams and the production of secondary raw materials has a local dimension which we cannot display in the in the in relation to primary raw materials that often we do not have. And of course, ahead the third headline objective that we had the third headline goal that we have is environment linked to the environment and the climate. Resources that stay in economy for longer are resources that are not extracted and processed and refined.
And there's a clear evidence, I don't have to to lecture this house on the fact that we would reduce CO2 emissions and GHG emissions greenhouse gas emissions from extraction and transformation. In terms of the process that we're facing, we have had a long and very structured consultation process, which yielded almost 2,000 contributions within the open public consultation and the call for evidence.
We had an SME panel, we had a targeted consultation. We had a number of opportunities to engage with this house in a number of initiatives that have been promoted by MEPs and groups of MEPs of different groups and nationalities, and this is really pleasant. We have produced a very weighty and complex impact assessment. We are facing the Regulatory Scrutiny Board tomorrow morning at 8:30, and we keep our fingers crossed for this. And the proposal is expected and is scheduled for adoption on Wednesday the 30th of September.
Thank you very much.
Yeah, thank you so much for updating us on the purpose again and on where you stand. So now we have the opportunity for the members to come in. And I have already Ms. Valean, and then Ms. Grosseman, and then myself. Please.
Splendid. Thank you very much, Chair. Thank you, representative from the Commission, for trying to update us. I have some questions in three areas of interest for myself. One would be the single market for secondary materials.
Orador 27
The Commission recognized already the fact that we have all these fragmented EPR schemes and waste transport rules. So, did I understand correct regarding the legal instrument, will the act contain directly applicable EU-wide end-of-waste criteria with mutual recognition, or will it once again delegate some of the harmonization to Member State implementation and comitology, which is how we got the current fragmentation? I hope I understood that we are aiming a high level of harmonization this time, to be clarified. A second area of interest would be regarding innovation and industrial development in the circular economy. If we look at the patent data, we'll see that most of the waste management and recycling technology patents filed at the European Patent Office come from outside the EU, with China alone accounting for a large majority.
And my concern is that prescribing too much, telling companies and scientists what categories of technologies to develop, will lead us to stagnation and not development. So this would be a question for the Commission: will you look at the technological neutrality in the upcoming act? And what would be the benchmark of success for this act?
Will you measure in obligations created or the patents filed? Because I think we are, we should aim at the latest. And the last area of interest or the question for today would be the WEEE and the urban mining. The Commission said that the WEEE revision will be part of the Circular Economy Act. So my question is whether it will be a market-building revision or a targets revision?
Because we had collection targets for years and I think we need now a business case. Recovering the materials from complex products is technically possible, but commercially less viable when the recovered content has no reliable buyer, no harmonized quality standard, and no price advantage over imported virgin material. So I heard you talking about the feedstock, but will the revised framework treat waste electronics as a strategic feedstock with harmonized recovery standards, legal certainty on secondary raw material status across all 27 Member States, and demand-side pull so that the materials recovered in Europe are actually bought in Europe? I rest here. Thank you very much for bearing with me.
Elisabeth Grossmann
Ja, danke Frau Vorsitzende. Danke auch für den Bericht, den Zwischenstand. Ja, der Circular Economy Act ist tatsächlich eine große Chance, wenn er gut ausgestaltet wird und könnte da wirklich vieles leisten, eben einen Binnenmarkt für Sekundärrohstoffe und recycelte Produkte aufbauen und auch natürlich die Nachfrage steigern nach recycelten Produkten. Ist aber weit mehr als Umweltpolitik, sondern auch ein industriepolitisches Instrument und wäre auch geeignet, Europas Abhängigkeit von importierten Rohstoffen zu verringern. Ja, und Sie haben es ja auch gesagt, natürlich strategische Wertschöpfungsketten zu stärken.
Entscheidend ist da tatsächlich die Ausgestaltung. Und deshalb möchte ich Sie fragen, welche konkreten Anreize geplant sind, tatsächlich den Anteil recycelter Materialien in Produkten auch vorzuschreiben, beispielsweise auch bei Ausschreibungen hier den Anteil von recycelten Produkten und Materialien vorzusehen, um das Potenzial auch voll auszuschöpfen. Und Sie haben von Vereinfachung und Vereinheitlichung nationaler Vorschriften gesprochen. Entscheidend ist hier natürlich auch, dass europäische Regulatorien ineinander verzahnt werden. Inwiefern ist das hier auch geplant?
Da gibt es ja viele Rechtsmaterien, die hier betroffen sind, natürlich der digitale Produktpass, die Verpackungsverordnung, Batterieverordnung, ganz, ganz viele Regulatorien. Mir ist dabei aber auch wichtig zu betonen, dass es keine Qualitätsverluste und keine Nivellierungen nach unten geben darf, weder was Umweltstandards betrifft, noch auch was soziale Standards betrifft.
Sie haben hier auch diese Dimension angesprochen. Es entstehen neue Arbeitsplätze und hier ist darauf auch Wert zu legen, dass das qualitativ hochwertige Arbeitsplätze sind, auch gut entlohnte Arbeitsplätze mit sozialer Absicherung und dass hier auch mit Sozialpartnern zusammengearbeitet wird. Inwieweit ist das auch gewährleistet? Insgesamt wirklich ein hohes Potenzial, das hier schlummert, ganz ganz wichtig im Sinne auch der europäischen Souveränität und einer nachhaltigen Wirtschaft. Dankeschön.
Anna Cavazzini
Thank you so much. Then I put myself on the list. Yeah, I can share a lot what the colleagues have said and also a lot of the goals you mentioned. I mean, for me, circular economy is really a no-brainer. It's the economy of the future.
It helps our sovereignty. It helps our CO2 emissions and the reducing the footprint, the material footprint. And it can also create and it is already creating new jobs. But of course, if you look at the numbers, I mean, it depends which study you take, but there's we're at the moment probably in a 10% rate of circular economy when you look at where we stand. And this is very, very low, unfortunately.
So I have the feeling, I mean, phrasing it more positively, there's a lot of potential. And the Circular Economy Act hopefully will unlock this potential if well done.
This is also what some of the colleagues have said. So probably my first plea is just to have really something a bit ambitious that you come up with something that will really uptake the circular economy in the European Union. And a little bit the more specific questions, when you look at the ESPR schemes and the harmonization, how do you want to make sure, and Elisabeth actually talked about it, that it's no race to the bottom because there is also well-functioning systems, other Member States have not so well-functioning systems. So how do you want to harmonize and not destroy the well-functioning stuff? Are you also looking at eco-modulation?
Because of course, this is also what a lot of, you know, companies tell us who are sustainable, that they need eco-modulation in order to really also have an advantage or like a fair competition with the other companies. And are you also looking or preparing overall targets? I mean, I know there's a big discussion if targets really make sense, but nevertheless, I have the feeling it is also clear incentive or like investment signal and a clear roadmap if you put concrete like circular economy or recycling or whatever circular targets in the law. These were my questions. So I give you the floor back for replying.
Thank you very much. Thank you for the questions.
Stefano Soro
I would like to start by saying yes, that we want to be ambitious. There's a lot of potential as you said. The circular material use rate, which is the sort of high-level, pretty crude indicator which was used also in connection with the launch of the idea of a circular economy act, is stagnating. It was 10.7% in 2010. It was 12.2% in 2025, a bit of a damp squib.
And this addresses also the question by Ms. Walen on what is the benchmark of success. The stated ambition was, is, the ambition stated by the President was to double this rate by 2030. To do this, we need to act ambitiously and act fast. Going into the specifics of the questions on the end-of-waste criteria, we want to couple the development of European end-of-waste criteria, but we know that speed has not been the middle name of the process, if I may say so, so far.
So we want to couple it with a procedure for mutual recognition between Member States of existing criteria. By the way, a disclaimer: we are we don't even have an approved a passed impact assessment, so everything I say, take a pinch of salt, there are no final decisions on any policy specific policy option. But obviously work has been going on for for for a while. On the positive side, we are certainly not going to prescribe any technical solutions in particular, so technical neutrality is obviously a feature of our intentions, as is, for instance, we are considering eco-modulation of fees. We will see.
We're not considering targets for the time being. As for the ESPR and in general the the remark by Ms. Grossmann, which is very pertinent, on the interplay with other regulatory and non-regulatory acts, at the beginning of our work we did a table, we came up with 27 adjacent, mostly regulatory but also policy-related acts with which the the future Circular Economy Act has a frontier.
So ambition and consistency, avoiding overlaps, avoiding over-prescription, avoiding to make enforcement on internal market and also at the border of internal market difficult or impossible is something that we are putting particular care in. The ESPR, which the chair mentioned, is one of these initiatives which has a clear frontier with the upcoming Circular Economy Act. It's based on a model of undisputed success, which is ecodesign and energy labelling. It's structured in a way which is mostly specific on products. There's a product-specific dimension which the Circular Economy Act, be they final or intermediate products—we're working on steel, next year we're going to work on aluminium, etc., etc.—doesn't have.
But there's a complementarity because obviously we are looking at waste streams which have a criticality dimension in the framework of the Circular Economy Act that will integrate, I think, in a coherent way with the work that we are doing with the ecodesign for sustainable—the implementation of the ecodesign for sustainable product regulation.
The social dimension, as I said, is important. We have a look also at some demand-side measures to create markets for secondary raw materials. Often it's the case that the difference in price between virgin or primary raw materials and secondary raw materials does not reflect the externalities of virgin raw materials. So we need to nudge a bit the business case for secondary raw materials. And we are thinking about a number of demand-side measures, in particular, and this is also talk about interplay, as regards public procurement.
For instance, some Member States have a blanket prohibition to procure second-hand or refurbished goods. Any of us who has in the family or is dealing with anybody between, let's say, 15 and 30 knows that they procure electronic devices on the back market, on the secondary market, and these come with a guarantee and work perfectly.
So there's inspiration to be had there. On patent, we're not looking at intellectual property in particular because we're not prescribing any particular technology here. But as I said, the benchmark for success will be an increased circular material use rate and an increase in the economic value of the circular of the sectors which adopt a circular model. The circular economy is not a sector, as is sometimes depicted, but it's rather an economic model that applies across sectors. Employment, sustainability of the social dimension of the of the employment linked to the circular economy will be the benchmarks for the success.
Thank you very much.
Anna Cavazzini
Thank you so much. Then I guess see you or a colleague again when you then present the draft report. So we're waiting for it. But thanks so much for the pre-information and also probably you can take some of our comments on board when when drafting the law and good luck with the Regulatory Scrutiny Board. This concludes agenda item 11.
If there's not any other business, the next meetings are, as you know, on the 2nd of September and then on the 10th of September. And I wish you a very good because now everything is over summer break. Relax.
Faça uma pergunta sobre este vídeo — analiso a transcrição para lhe responder.