Trustworthy AI in Public Services: Transparency, Accountability, and Crisis-Resilient Communication – MT 03 2026

so let’s pick up where we left off and hearing the bell feels like you’re in the theater so the show is starting um this is the final stretch of our event but by no means the least because we have another two main topics to go through and go a little more in depth into some of the topics that we have already touched upon as is the case for our next session we’ve been talking a lot about trust trustworthiness whether in internet governance or in all the related issues and topics and this is what we’re now going to dig a little deeper into the new main topic we’ll look into how we can ensure a trustworthy ai in public public services, a key issue definitely that has already been mentioned by several of our speakers or by some of you in the room as well.

So to take us through this very discussion, I’m happy to hand over moderation to Gabby Scucciate. She’s the CEO of Vilnius Business College and her co -moderator, Aisha. Welcome.

Hello, good afternoon. It’s a very good day and a very good time for the session because it’s after lunch, so I think you feel relaxed. And we will talk about a very important issue, about trustworthiness in AI. Europe is built on democracy. So democracy was built. It’s born in Europe. But democracy is not only what we inherit, it’s what we are obliged to keep it. And we live in the times when democracy is relived and recontextualized in new contexts, in new social, digital, political contexts. And one of these contextual measures, of course, is AI. And the context of this session is built upon trust. As today, we live in the world where AI is intervening a lot of the dimensions of social, political life, but also public services.

And public services are a very sensitive area where every person, individual, is… is having interaction and is affected by decisions made by government or the public bodies. So today we will discuss who we can trust and is AI trustworthy and trustable. So I’m very glad and honored to be moderator of this discussion with my co -moderator, Aisha. And now I’m giving floor to her.

Thank you so much, Gabija We thought this would be a great idea as one -and -a -half -hour session might be difficult for one person to moderate. Just to touch upon what Gabija was saying and perhaps to add on that, we know that the use of AI and more specifically the use of AI in public services raise serious concerns. When decisions that affect citizens are supported or shaped by algorithms, we must ask who is responsible for those decisions and how to ensure to make decisions. sure while we’re pursuing efficiency, it does not weaken democratic control, human dignity, or the right to good administration. And this becomes especially prevalent when we’re talking about times of crises.

In crisis situations, governments may rely more heavily on digital tools, automated systems, and rapid communication channels. But precisely in such moments, we should remind ourselves that transparency, a word that perhaps we’re hearing way too often today, should not be treated as an end in itself. Transparency must serve accountability, equality, and access to redress. And from an equality perspective, some of the harder questions are whether regulators, equality bodies, researchers, or affected communities can see who is targeted, who is excluded, who is misrepresented, and who bears the cumulative impact. And this is why we’re… I’m very interested in this topic, and as I think one of our speakers will delve into it a bit later, we work with equality bodies within the Council of Europe who independently oversee the principle of non -discrimination and whether it’s respected by the use of AI systems.

And they report to us that citizens oftentimes do have a suspicion that they’re interacting with an AI system, but cannot be certain if they have been doing so. And this impacts their access to public benefits, court procedures, any other sectors of public life. And without understanding whether they’re interacting with an AI system, individuals cannot exercise and enforce their rights. With this, I’d like to give the floor back to Gabija for the session format.

So there’s a lot of concerns for the beginning. So for that, we have a very distinguished panel of speakers, and we have one speaker online, as well. So the session will be organized. This way that we have four speakers, they will make their presentations. And then the most important part will be engagement of you as a public to hear your voices and to comprehend on some cohesive, you know, like remarks or minutes from this session, which will be finalized as our messages of this working group. So I want to start introducing our distinguished speakers today. Dimitri Gugunava, Ministry of Justice of Georgia. He’s also, he represents Digital Governance Agency. He’s a head of Digital Governance, Cybersecurity Strategic Planning and Analytical Unit.

And also. The Council of Europe representative to the Steering Committee for New and Emerging Digital Technologies. That’s a lot in one person. I hope I didn’t miss something else. So welcome to the stage. I’d like to introduce the second speaker, Nele Lorokins. She’s a project manager of strategic litigation and AI at Equinet, the European Network of Equality Bodies. And Nele has over eight years of experience from a national fundamental rights authority leading work on emerging technologies on non -discrimination. And previously, Nele has also chaired the European Network of National Human Rights Institutions Working Group on AI and represented ENRI at the Council of Europe’s Committee on Artificial Intelligence. And as I said, we have one speaker online, Professor Ebba Ossian -Nielsen.

She’s an expert in open and online learning, and she won several international awards, like Open Education Global Award for Open Leadership. She is also a professor at the University of New York. also a honorary member of the ICDE, where she also serves as a board member and contributed with several research projects. That organization is International Council for Open and Distance Education. And our last speaker today, Mr. Jaroslav Ponder, is the head of the Office for Europe at the International Telecommunications Union, ITU. Representing ITU in Europe and directing actions, projects, initiatives, and expert groups targeting 46 countries in the European region. This made me also remember that I forgot to introduce you our guiding questions that our experts will be touching upon today.

Our first question is, how can public services balance efficiency, transparency, and inclusivity when deploying AI while maintaining human oversight and democratic control? Our second guiding question is, what role does ITU play in the development of the world’s most advanced technology? can anticipatory governance, civic participation, and AI itself play in identifying and mitigating risks such as bias, exclusion, and unequal access? And our last guiding question is how can current regulatory approaches be adapted to address real -world implementation challenges, especially for vulnerable or underrepresented groups? So the floor is ready for the first speaker. You may take this stage or from there. It’s up to you. So, Dimitri Gugnava from Georgia.

Yes, sure. We can. Thank you so much. It’s an honor for me to be here. Thank you so much for the invitation. If anyone has raised the expectations, for my long title, please lower the expectations because that’s the secret for being happy. Before I begin, let me make a usual personal safeguard. I’m speaking in my personal capacity. Everything I say represents my own views and not necessarily the position of the Council of Europe or the government of Georgia. It’s really a pleasure to be here. And here I joined yesterday when the topic of discussion was this new technologies and intangible and how this interconnected all these topics are. The trust, the public services, the human rights, democracy and the rule of law.

I shall try to contribute here from the perspective of the Council of Europe, especially if this microphone also works, I can. I can please before it does okay uh from the perspective of the council of europe and just recently established the committee on new and emerging technologies cd net uh which is responsible for popularization of the uh framework convention on ai uh human rights and democracy and the rule of law so we we mention very frequently the the title of the of the of the session uh mentions to towards the transparency and accountability and i think we must ask the fundamental question why do we need to safeguard these these values the the one ultimate answer to to this is to build trust and why do we need trust because the society works on fundamental principles and one of those is the trust as a principle and how the trust is built the build the trust is built by by meeting the expectations.

And what are the expectations? The expectations are we see the international access on human rights, including the European Convention on Human Rights. In public services, the trust is foundational, and it’s foundational for relationships between the citizen and state. And when a citizen is a user of the public service, he or she is not just a user. He is a rights owner, and the service provider, public service provider, is not just the company, let’s say, who is providing services. It’s also bound with the obligations, obligations coming from the law, from the national laws, from the international laws. So I want to move to describing the situation, the situation. And I… While talking about AI, different from many existing and previous technologies and technological revolutions, AI is not just amplifying the human force, labor force.

It does make us just faster. In some cases, it can generate the content, it can make decisions, it can classify people, recommend choices, and influence or in many cases even automate the decisions that are having the effect on real life. This is not happening in digital world only. So these decisions, these classifications have the implications on real lives. And I think this is very important to mention. In the public sector, this creates quite real risks. AI may help. citizen to receive a service faster or reduce the cost of the public service delivery. But at the same time, these systems, these technologies may also wrongly classify that citizen is ineligible. It may help to detect fraud, but it may also repeat historical bias.

It may help public authorities communicate during emergencies, but it may also spread confusion if information is inaccurate, inaccessible, or not properly supervised. At the Council of Europe at CityNet, we are having discussions on many topics, and one of those is the data itself based on which these systems usually work. And in some cases, we have examples when the data by its nature is biased, and how can we ensure that the systems working on this data are not biased? And how can we ensure that the systems working on this data are not biased? And how can we ensure that the systems working on this data are not biased? I would like to group these concerns from my personal experience into three groups, which are related to the using of AI technologies and AI -based technologies.

The first is the most common threat and maybe the most simple one, the misuse by bad actors. Let’s say bad guys are misusing the technologies to generate illegal content or have different kind of achieved some illegal goals with using these technologies. But this is the simplest one. The second type of AI -related risk is the loss of control or insufficient control. You may remember from the fantasy movies, which are already documentary ones, when the AI systems are out of control. And people… can only stop the electricity to the system. So when the system is out of control, this is the second type of risk. And the third one, and I think the biggest one, is the structural over -reliance.

AI may become and maybe has already become so deeply embedded in the process of public administration that institutions depend on systems they cannot fully explain or control. This comes to not only on the organizational level, but also on the individual level. The public servants who are making decisions, who are doing their job on a daily basis, are using AI systems. This is especially important in high -impact areas. such as social protection, health care, education, justice, policing, digital identity, and so on and so forth. So here comes the question, after identifying the risks, how should we respond? Because the risks may have an impact, if they realize, on daily lives of ours, also the future generations.

Ethical guidelines on AI are very valuable. They are very important. The process of elaborating these ethical guidelines have already led to identification. Real risks led to very good discussions. But the question is whether these ethical guidelines and non -bounding regulations are enough to respond. This doesn’t mean that regulations should be used to identify real risks. They should be used to identify real risks. They should be used to identify real risks. They should be used to identify real risks. They should be used to identify real risks. They should be used to identify real risks. They should be used to identify real risks. They should be used to identify real risks. Food regulation can at the same time support innovation by creating trust, legal certainty, and clear expectations for citizens, for public authorities, developers, and the deployers of AI systems.

This is especially important in the digital environment where borders often have limited practical meaning. We now know how much European Union, for example, is doing for unified digital market, how much different countries are doing to unify their digital environments by ensuring the interoperability so that citizens of one state can receive the very same quality public services in another state. So not only these technical things like interoperability. and technical, legal, technical, and semantic levels are enough, but also We must be sure that the public services are delivered in the very same ethical and fair way and transparent way on the national level as well as in the international level. So the aim of the – so this is the argument why we need the international regulation.

Not so that national level regulations are never enough. So the aim of the international regulation is to define common principles and safeguards, human dignity, transparency, accountability, equality, non -discrimination, privacy, human oversight, access to remedies, and protection of democratic processes. In simple terms, technology must never operate in isolation from human judgment. And legal responsibility. All these principles that I mentioned is part of the framework convention. on AI and human rights and democracy in the rule of law. This is where the Council of Europe response becomes very important. The Council of Europe didn’t start from the zero. The Council of Europe already had some international conventions which were somehow covering these technologies, AI technologies. But the decision was made to make the framework convention specifically on AI.

All started with the CAHAI, Ad Hoc Committee on Artificial Intelligence, which has mandate to measure the feasibility, to assess the feasibility. And the CAHAI concluded that the Council of Europe should move towards the binding international instrument on AI supported by practical guidelines. And the CAHAI concluded that the Council of Europe should move towards the binding international instrument on AI After that, the feasibility study was conducted. Council of Europe established Committee on Artificial Intelligence which had a mandate to negotiate the Framework Convention. The process also involved not only the member states of the Council of Europe but also countries which are not member states of the Council of Europe. Also included civil society, academia, the private sector and international organizations.

I think this is very crucial since we want to have the legally binding document which can work on the international level. How many and how diversified the involved personnel was in the process. The result of the CAI was the Council of Europe Framework Convention. which was adopted in May 2024 and was opened for signature in Vilnius later that same year, 2024. It is the first legally binding international treaty on AI with global reach. This convention, since we have the auditorium consisting of students, non -legal professionals, I want to emphasize here that convention should not be confused with the EU AI Act. The EU AI Act is a detailed regulatory framework within the European Union. The Council of Europe Convention is the broader international treaty on human rights, democracy, and the rule of law, and it’s open beyond the EU.

By the way, since I mentioned the EU just 10 days ago, maybe nine days ago, the European Union also finalized. The process of ratification of the convention. European Union is one of the signatories of the Convention, so this means that no EU member states independently will sign the Convention, no EU member state will ratify the Convention, but beyond the EU, there are also very important players in the market and countries who supported the process and also the countries who have very huge impact in this domain, including United States of America, the United Kingdom, Japan, and many other countries, including the Georgia that I’m representing, that I’m coming from. The Convention creates a common legal ground for responsible AI governance.

It is based on key principles such as, as already mentioned, human dignity, equality, non -discrimination, privacy, transparency, oversight, accountability, reliability, and safe innovation. But it doesn’t stop at principles. Even though it’s technologically neutral, it also requires practical safeguards, including access to relevant information, the possibility to challenge AI -informed decisions, compliant mechanisms, procedural safeguards, and effective remedies. It also requires risk and impact management throughout the AI lifecycle in simple terms. Risks and human rights, democracy, and the rule of law should be identified, assessed, prevented, and mitigated. This is where the Hugh Darrier document comes in, which is not an integral part of the convention, which is not a legally binding document, but still plays a huge role in terms of implementation of the convention.

Hugh Darrier helps to translate the convention into practical questions and steps. Hugh Darrier has four main elements. The first one is understanding the system and its context, which is so -called COBRA. involving affected people, assessing possible harms, and taking action to avoid or reduce those harms from the beginning. All these documents are publicly available on the website of the Council of Europe. So if anyone is interested, you can dig down the documents. Now let me move very quickly, since I’m out of time, to move to the next part and try to connect this directly to public services and to the guiding questions of this panel. So the first question was, how do we balance efficiency, transparency, inclusivity, human oversight, and democratic control?

Efficiency is very important. Efficiency matters. It’s a public interest to have governments who are efficient. But it cannot be the highest value. It cannot be the highest value of public administration. A public service that is fast. but unfair is not truly efficient. AI in public services should be judged not only by speed or cost reduction, but also by legality, fairness, accessibility, accountability, and human dignity. Second, how we identify and reduce risks such as bias, exclusion, and unequal access. Qdaria helps that because it looks not only at the technology, but also at the context, the data quality, historical bias, institutional capacity, accountability, and the situation of affected groups. Third, if we shall have time to move back to the questions later, then I shall just make a conclusion and finalize my word.

Let me conclude here with three very short, but I think very important, messages. The first one is AI in public services should be judged not only how much time or money it saves, but by whether it strengthens or weakens trust between citizens and the state. Second, human oversight must not be reduced to a checkbox. It must be meaningful, competent, accountable, and capable of changing outcomes. We have that kind of experience from the GDPR, like using the simple language, right? And the third, the people who are most affected by AI systems must have a voice in shaping these systems. The faster the car is, the more reliable the brakes it must have. The

thank you so much for your very informative and insightful presentation is my mic on okay so let’s turn to the second presentation yes now it’s working thank you very much for the presentation I’ll move on quickly to Nele Nele the floor is yours if you want to sit there or go over to the stage whichever you prefer

Thank you very much, Ayça, for that kind introduction. So, good afternoon, everyone. I am delighted to be here. As you heard in the introduction, I am working for Equinet. Equinet is the umbrella institution, Brussels-based, of all equality bodies. We have 48 members, European equality bodies, apologies. And what are equality bodies? Those are public but independent bodies that have the task to promote equality and combat discrimination, and they have very specific powers to do so. You might also know them under the name of National Human Rights Institution, Ombudsman or Ombudswoman. So these are equality bodies. Today, my presentation will focus on three main points. So the role of equality bodies under the new legislative framework works to have been addressed already, the EU AI Act and the Council of Europe Framework Convention on AI Human Rights, World of Law.

Secondly, I will share some best practices in improving and addressing trustworthy use of AI by public administrations. This is a project that has been carried out in cooperation with the Council of Europe and the European Commission. And thirdly, we will focus a little bit on something more technical, but very important in this context, and this is the standardization process. So before delving into those three points, I will set the scene. So all of you in the room might be familiar with the Dutch child care benefit scandal or the Amazon hiring recruitment tool. You have heard probably of a lot of… potential cases of algorithmic discrimination, yet it might very well be that you have not personally experienced it.

Also, in the introduction already, the Council of Europe kindly indicated that even though people know this is happening, equality bodies are receiving approximately 10 ,000 complaints on a yearly basis, depending from institution to institution, but individuals are not identifying it as such. And what are two of the reasons, therefore, we would call them information asymmetry and power asymmetry. Information asymmetry or information gap can be defined as a significant imbalance between access to, understanding of, and control of information. Simply put, you do not necessarily know whether or not you are being shown a job or housing advertisement. And even if you do know, you do not necessarily know how the system, based on input A, decided output be?

And this relates to the threshold problem. So even if you realize that you have been victim of targeted pricing or your feed is showing you in some extreme left or right content, how will you address it? Would you have the necessary knowledge, financial resources and even energy that you could dedicate to assess this and find avenues for effective redress? So these are a little bit of the complications and this also illustrates the importance of equality bodies. So equality bodies have this mandate to promote and protect individuals both in the public sector as well as in the private sector and the legislation, the EU AI Act and the Council of Europe Convention both address specific attention to cooperation between market surveillance authorities and existing authorities protecting fundamental rights.

Because it’s important to highlight the importance of equality. The AI Act, 150 pages, very complex, technical, without annexes. The Council of Europe Convention, only 12 pages, so easy to read. I recommend you to do it. So both explain the means of cooperation between these two fundamental actors, and this relates to equality bodies have to face the lack of meaningful transparency, the complexity of the systems. They do not necessarily have the required technical expertise to conduct investigations, etc. So within the AI Act, they will have now specific access to documentation rights, the right to trigger testing, but also more importantly, to counter a little bit that information asymmetry, market surveillance authorities will also need to inform them when they establish that there is a serious risk to fundamental rights or potential risks to fundamental rights.

So this is a little bit more about the role of equality bodies. And now you might wonder, okay, are they now ready to… rise to the challenge. And here comes in a little bit of the best practices that I will demonstrate to you. So I’m very proud to present this TSI project. The commission people here might be familiar with that very specific term. It’s a SG reform project that three of our members, Belgium, Finland and Portugal, are the beneficiary authorities of this project. The project is being implemented and co -funded by the Council of Europe. And today I will present you three outputs of the project that are extremely important when we’re discussing the topic of today.

So first, new legislation. AI Act, I told you it’s long. The Council of Europe Convention, how does it relate to the AI Act? And existing anti -discrimination legislation. We know there are these EU directives, but you also have national legislation that goes often beyond? So more protected grounds, intersectional discrimination, multiple discrimination. How do all these things relate to each other? How also is it possible at this moment with this new legislative framework to address new forms of discrimination, for example, through randomized formation of patterns and proxies or inferred characteristics, even if you try to make a system protected characteristics free? So is that possible? Would you like to have an overview of the legal protection against algorithmic discrimination in Europe?

The current frameworks and gaps you will find in this booklet. Secondly, the AI Act, although it’s essentially a product safety approach, it also tries to protect against fundamental rights and address fundamental rights harms. But which provisions are most key? How do they relate to existing legislation? Like the GDPR and what role could be there for equality bodies or public interest organizations? All these relevant provisions you will find in this second booklet. For example, some of you might know there is an annex tree detailing high -risk systems. High -risk systems is considered by a lot of people like the core of the AI Act because the majority of duplications are applicable only to the high -risk systems.

This annex can be updated. How can we make it meaningfully participatory? You will find all this and more in that. And third, last but not least, in September, the Council of Europe and the Commission will publish a methodology for assessing AI -related discrimination cases. Which information can you ask from the deployer? Which information can you ask from the provider? How do you assess this discrimination, this information, in order to conclude whether or not bias, can also mean unlawful discrimination? This brings me to the third part of my presentation. Quick time check. Two minutes. Okay. Thank you. So bias is a very central word in discussions on discrimination and AI. Bias is not defined in the AI Act.

Bias, there are a lot of interesting examples in the explanatory reports of the Framework Convention. It’s important to highlight that there is social forms of bias and also technical forms of bias. And both need to be addressed. The social, like, for example, automation bias, looking at it from a more socio -technical point of view, taking into account for historical patterns of discrimination, existing societal inequalities. This we can do via meaningful transparency, fundamental rights impact assessments, and so on. The second part, the more technical part, bias detection and correction, is also addressed via the standardization process. And I’ve mentioned a lot of times AI Act and to a certain extent Council of Europe Convention approach these issues via a product safety approach.

Product safety is measured through technical standards. These technical standards for the EU are currently being developed in the Joint Technical Committee 21 from SENS -SENELEC. These, to the people that have experience with the standardization committees and maybe will learn later, are not necessarily human rights experts in those kind of rooms. So it is very important that those institutions that will now decide what is the acceptable level of residual risk or what kind of bias can we accept, that there is also some relevant fundamental rights and equality expertise in that. And there I’m very also. I’m pleased to announce that Equinet, since 2023, has liaison status at SenSanElec GTC21 and has been trying to contribute that fundamental rights point of view in the discussions.

Due to a technical hiccup, I’m unable to share the QR codes of all of the publications I’ve been talking about. Please come see me after the break if you want more information and also work on GTC21. One is a little bit too technical. I need to discuss here. So I will finish with a very nice best practice from the Council of Europe, RIDERIA, and this is something that’s called Zero Questions, and this would be an answer to the three guiding questions of today. And the zero questions, actually, especially in the public sector, where we can expect a heightened level of scrutiny and taking into account the risk and assessing avoiding them, when we’re using AI, is asking…

yourself the question whether the use of AI is appropriate in a certain situation, whether non -automated alternatives are even considered, and if you cannot achieve the required level of transparency, for example, or legality, then you should not deploy the system at all. These very specific guidelines, I refer you to the Hedaria. So thank you very much for your attention.

Thank you very much, Nele, for your very interesting presentation. I would like to remind our speaker that I know you have a lot to say, and thanks for the very interesting presentations, but to also give time for the interventions. We are limiting them to 10 minutes straight, and I give the floor to Gabby.

So let’s move forward. And online, we have Professor Ebba Ossiannilsson. I already announced all your… credentials, so I will not repeat them, and giving you this virtual floor for your presentation.

Thank you so much for the kind invitation to be with you today and to contribute with some insights. It’s a great pleasure and honor to be here with you, although it is online, it is very special. And thank you to my co -presenters who have presented very, very, very insightful contributions. I’ve been told I can’t share any slides, but I will read them for myself. So we have those three guiding questions. How can public services balance efficiency, transparency and inclusivity when developing AI while maintaining human oversight and democratic control? Second, what role can anticipatory governance, civic participation and AI itself play in identifying and mitigating risks like bias, exclusion and unequal access? Third, what role can public services play in identifying and mitigating risks like bias, exclusion and unequal access?

Third, what role can public services play in identifying and mitigating risks like bias, exclusion and unequal access? How can current regulatory approaches be adapted to address real -world implementation challenges, especially for vulnerable or underrepresented groups? So I will start with a summary. So you have that already from the beginning, what I will talk about. First of all, a trustworthy AI is not only about safe technology. It is about preserving democratic legitimacy and human agency, inclusion and public trust in times of uncertainty. So public AI systems is part of critical societal infrastructure and similar to health care, education, welfare and civic communication. To take the whole ecosystem on and also very much stress about humanity. And also about well -being.

So for the first question. A trustworthy public AI system must always leave room for human judgment, human dignity and democratic accountability So thus, we need to advocate for, and when I say advocate, I mean all of us It is not just one person or one authority or one special office or regulation or whatever We are all part of this game and we need to advocate for mandatory human in the lead governance for high impact public decisions And I have stressed here not just to be humans in the loop, but to have humans in the lead And also about algorithmic impact assessments and public transparency registers Question number two Governance must improve Move from reactive regulation towards anticipatory with governance And public institutions should identify risks before harm occurs.

So, thus, we all need to advocate for that trust can’t be engineered afterwards. It must be designed into systems from the very beginning. A human -centered, anticipatory, governance, and resilient civic AI ecosystem. So, you see, I stress very much about the ecosystem and to have this holistic approach to secure that nothing will fall in between the chairs. Because things are related and interconnected to each other. Question number three. Trustworthy AI is not achieved when revelation is written, but when fairness, accessibility, and accountability, and accountability, and communication are experienced by citizens in practice. I started with a summary so my conclusion is trustworthy AI is not achieved when revelation is written as I just said, but when fairness, accessibility and accountability are experienced by citizens in practice and for that, trust as relational and societal is important and we also need to work on leadership through uncertainty because uncertainty is the only certain thing we know exists and the need for inclusive, sustainable and human -centered AI ecosystems rather than purely technological optimization so that was my very brief contribution for those 10 minutes thank you so much

thank you professor for this concise but very impactful presentation which you delivered thank you so much and we hope you will stay with us online for the further questions

I will. I will be happy to take questions in case.

Thank you, Ebba. It’s my pleasure to give the floor to our last speaker, Jaroslaw.

Thank you very much for involving us in this panel. It’s our great pleasure to be here with you as representing the ITU, International Telecommunications, being the UN Agency for Digital Technologies. AI is challenging, but it’s involving not only the ITU, but a series of different UN agencies. And this is what we are representing currently under the brand of the AI for good, because we believe in the good of the AI. So that’s why I’m very much happy to see also our colleagues and friends from the UNESCO, with whom we are co -creating and also co -chairing the interagency taskforce at the UN level in order to make sure that the UN… UN acts as one in addressing the challenges related to the AI, building upon the specific mandates and the competencies of different UN agencies in order to ensure that the answer to those questions are coming in the comprehensive way, encompassing all discussions in the UN.

And this is something that relates to what we discussed at the beginning of this day when some speakers were calling for clarity and not fragmentation of different workflows. So talking about this, of course, we as the technical agency and the agency which since 160 years are dealing with the digital technologies accompanying to its sector. bringing the technologies to the people. Of course, the question is, what can we do for the sector? What can we do for the public sector? And, of course, our mission is to not only to connect the world, but also to ensure sustainable digital transformation, which brings us specifically to the issue of how to make transferable AI in public services, but also how to make sure that those technologies are not only the technologies, but they’re representing the human face and the human needs, addressing the human needs.

Of course, our core contribution here is coming with the development of the technical standards. As of today, over 400 standards are being developed, and once we are speaking, many others are being developed. We are under the development in order to make sure that we’re building upon that. And, of course, they’re used later in order to make sure that the countries and that the regional providers are building the public services upon those standards. They’re building the procurement processes, and they’re making sure that this, what has been designed as the global understanding on the technical issues, is then meeting the citizen. Of course, to make sure that certain concepts are reflected properly, it’s not that easy. I think our colleagues were already mentioning this is a huge challenge.

And so, once you’re speaking to the group of the engineers who are coming to Geneva, who are coming to the different parts of the world speaking the technical jargon, to understand why they should care about this human -centric approach. It’s not understood. We started this process several years back, so we have done significant progress already in this, and we’re pushing, and thanks to the cooperation with the Council of Europe, support of the European Union, strengthening dialogue with the European institutions. I think we’re building the perception that the future of the standardization cannot continue without including the human -centric approach for digital transformation. And this happens. We’re progressively seeing the acceptance of getting involved the engineers into the discussion on the human rights -based approaches, on the human -centric approaches, on human rights, demystifying the concepts, and also clarifying what, in fact, is the impact of their action.

And I think this is important to understand that the simple standard, which sometimes is 15 pages, has such a big… big impact on this, what is happening at the national level, that we have to imagine this. And this is the reason why also in the same time, in parallel, we’re working with the policymakers and regulators to see how to assess their readiness to accept those standards, how to implement them at the national level, and how to understand them also at the national level. There’s a significant demand for the capacity building. We have launched several trainings through the ITU Academy, which is reaching out to the over 130 ,000 professionals in the United States, and the numbers are growing.

Several courses are delving directly to the question how to make the AI in public services really trustworthy and accountable. And and how to make this happen. of course this is only the dripping ocean once people are getting trained and very much very often they are going to the other sector the private sector so the more efforts and more of spreading the news is necessary and therefore also apart from the taking a stock where the countries stand in terms of the implementation of certain concepts in context of the AI is of the high value. Recently we have released the second update of the readiness framework for the AI assessment we are kindly inviting you to take a look at this this is a great exercise to digest what is happening at the international level and how to compare the countries in this sense but what I would like to also to mention that that is And still we are facing the challenge of the level of the digital skills in our countries.

I think most of us will be shocked when we will be hearing that only around 100 countries do collect the information about the digital skills in their countries. In other more than 90 countries, there’s no understanding where those capacities are. So you can imagine that those citizens who are exposed to the digital services, in particular AI, they will be excited about using everything. So I think we need to also take this as the comprehensive approach towards the rollout of the AI services, in particular to the citizens, with the potential… with the potential of the scale -up and making sure that those concepts which are discussed here… and let me now speak as the head of the Europe Office more with the European…

approach, that those concepts are very well understood at the global discussions. I think we really need to make sure that the European voice is heard properly in the global discussions, that the concepts are explained in the way that is accessible to all other regions, that the dialogue is happening in a meaningful way, seeking the consensus and understanding why different concepts are developed there. And I hope that July meeting, which we already mentioned several times during these two days, July meeting of the global dialogue on the AI, as well as AI for Good Summit, will offer the opportunity to deepen this discussion and to get closer between the regions and making sure that such instruments like convention is something.

Not seen as the European product. But as the solution for the future, because I think we need to make sure that at certain moment we have the one single reference point for certain challenges which are related to all. Maybe not to all, just to conclude, and saying not to all, I wanted to say about our core business, second pillar, because 2 .2 billion people are still disconnected. So 2 .2 billion of the citizens of the world still didn’t have the occasion to enjoy even simple Google search. So we have to keep this in mind while we are developing sophisticated systems for bringing the government services closer to the citizens. On the other hand, we have to make sure that the exclusion is not I want to use the term tolerated, but it’s not accepted to the level which we are facing today.

So urgency is there. And we look forward to welcoming you in Geneva in July with the delegations at the highest possible level, but also at the expert level to make sure that the discussions at these three meetings happening together can be really fruitfully advancing the discussion on AI. And but also on the implementation of the WISIS plus 20 outcomes. Thank you very much.

Thank you very much, Yaroslaw. And also thank you for ending on a note where it’s a good foot for thought for all of us to think about digital exclusion and remind ourselves of that. I’d like to give the floor back to Gabija as I’m happy to move us forward to our guiding questions and interventions. Gabija, the floor is yours.

So the scene is set up with a lot of grounded information And our panelists really have this expertise in the field But now is the time to listen to your interventions As we want to hear everyone who is registered here We will try to be in time But it also depends on how much interventions will happen in real So the queue is set up according to who registered the first So the first is first registered, first served So the first question is How can public services balance efficiency, transparency, and inclusivity When deploying AI while maintaining human oversight And democratic control? And we were told not to wait too long So if the person is not appearing, we will shift to another one So we will start from Pari Esfandiari from Global Technopolitics Forum online.

Yes, she’s here. Parian?

Yes, thank you very much and for the opportunity to talk. It has been very informative and I want to thank all the speakers. My comments are focused on this question one. Some of the points I want to make may already have been mentioned or touched upon, but I would like to continue with them, not to repeat, but to emphasize and hopefully to expand. So I think one challenge in this discussion is ensuring that transparency and human oversight remain central governance principles rather than becoming purely procedural requirements. Public services, as many of you touched upon, are not commercial platforms. Their legitimacy depends not only on efficiency, but also on fairness, accountability, and public trust. So, yes, AI can improve efficiency, but meaningful human oversight must remain built into decision -making process from the start, especially in areas like welfare access, healthcare migration, or other services where automated decision can deeply affect people’s lives.

And there is also a structural concern I have here. AI is increasingly centralizing data, knowledge, and decision -making capacity within a small number of powerful actors. If public administrations become too dependent on these systems, transparency and democratic oversight may gradually weaken. And this is why public participation and multi -stakeholder governance are essential. Citizens should not only be subjects of AI governance, but participants in shaping how these systems are designed and deployed in public life. And this cannot be treated at one time.

Pari, the time is off. So, Pari, thank you so much. We will not be answering the questions or interventions. Now we’re just compiling these insights, and then the experts, if there will be enough time, will make their remarks or, you know, just comment on something. Thank you. And let’s move to Aduna Nechomolato from – no, she’s not here, or he. Then Adriana Rodriguez-Novo from – Fundacion Galicia Europa. Europa. It should be here. No, let’s move to Sandra Martike. Sandra Martike. Sandra Martike. You are here? Yes, Sandra, please.

Okay. About this question, I think that what is really important for us because we are a company, a Swiss company, and this is really the partnership, partnership like public -private partnership because, you know, technology for technology is not what we want to keep in mind. We want to, our mission is really to bring the light on The problems, how to solve it, and the AI will help. And will help in a very protected or measured way because we know the power of AI. But finally, we are the master. AI is not our master, right? And also keep in mind all the regulations like GDPR, privacy, and everything. And this is the first step. But we also need to involve the people, the citizens.

Because most of the time what we do is that we are between us. So B2B, so private companies, governments, and the voice of the final beneficiaries is not there. It’s not heard. It’s not heard. And, yes, this is what I wanted to contribute to, and really the partnership, I think, is really important. And this is what I see here in the European Commission, that there is a lot of collaboration between people, institutions, private sector, and also young people. I think that they are the, also they have their voice, and they also consider that AI is a really great tool, but it’s also a lot of risk that we need to mitigate.

Thank you, Sandra. Thank you for your contribution. And when I pronounce a person’s name, if you are in the hall here, please raise your hand, and it’s better to see you. Maciej Plasecki from DigitalPlanet. Digital Safety Advisor, Maciej. Actually, that one is online. It’s online. yeah and do we have him online we do but we had some problems connecting before so so maybe we can move on to the second one uh sunsitsa rosic from central european university and sorry if i pronounce your names or synonyms incorrectly sunsitsa no no camel el hilali from unesco camel no it seems jialin liao from

um thank you for the for i have briefly compared to governor’s um i have briefly compared to governor’s um i have briefly compared to governor’s approaches and then put forward three approaches and then put forward three approaches and then put forward three universal measures universal measures universal measures the eu’s ais and regulatory center the eu’s ais and regulatory center the eu’s ais and regulatory center bosses put human rights bosses put human rights bosses put human rights front and center adopting a sixth -generation model. In China, AI tools such as DeepSeq are widely deployed to improve administrative efficiency. While I do not endorse the entire governance model, I would like to draw attention to one notable practice.

While it tests precision systems that assigns clear individual accountability to officials for all decisions, this novel accountability framework is worth of reference. Both sides face shared challenges, upholding procedural standards, exercising administrative discretion appropriately, and guaranteeing full accountability. The following three measures are possible across jurisdictions. First, structured challenges. Second, transparency. These decisions must be accessible and as opposed to safeguarding cultural integrity. Second, meaningful human review. Any automotive output that impairs people’s fundamental rights and interests requires rigorous, contest -based assessment, never a mere ball -sticking exercise. Third, clear liberty. Every AI -assisted administrative act must be attributed to a desired officer, with the government undertaking ultimate guaranteed liberties. By comparing Europe’s right -first philosophy with its rigorous individual accountability mechanism, we can build a governance model that balances operational efficiency and public oversight.

Ensuring public sector AI remains transparent, inclusive, and legally compliant. Thank you.

Thank you so much. Thank you for your contribution. Mikita Danilov. Are you here? No. Mariam Japaridze from YouDig.

Good afternoon to everyone who tries to tackle with AA trustworthiness while still trying to keep the human oversight and democratic control. My name is Mariam Japaridze and I’m a YouDigger. First things first, how do service providers gain public trust? By staying consistent and true to their words, which sounds easy, right? But reality is harder, especially if you want to balance consideration and compromises we all as a society need to make for successful coexistence. I come from Georgia, a country where trust in public institutions and services have often been connected to questions of democracy, inclusion and resilience. Keeping these values is a challenge for every government and adding AI to public services, which seems inevitable, makes processes even harder.

AI can keep public services under control, respond faster and improve access to services. But we know that public institutions are different from private companies. Their purpose is not only optimization, but fairness, transparency, and public trust. And trust cannot be automated. Citizens must still understand how decisions are made and have rights to challenge them. Otherwise, efficiency risks becoming something that distances people from democratic processes. Also, I think inclusivity must become part of AI design from the beginning, not an afterthought. As for my suggestion, perhaps the goal should not be to create AI -driven governments, but governments that use AI while remaining visibly human, accountable, and democratically controlled. Because ultimately, citizens should never feel that public services are speaking at them through algorithms instead of listening to them as people.

Thank you.

Thank you so much, Mariam, for so much well -articulated contributions. Giovanna Deditz. Giovanna no not here and the last one in this guiding session is Florian Roussel from European Pirates Civil Society Florian doesn’t seem in the whole ok so thank you for the contributions to the first guiding question and let’s move to the second one t

hank you Gabby and now moving on to our second guiding question what role can anticipatory governance, civic participation and AI itself play in identifying and mitigating risks like bias exclusion and unequal access I’ll also go start with the line right away Axel Mazolo nope ok Ranyan Timusina I apologize also for the pronunciation I apologize also for the pronunciation I apologize also for the pronunciation I apologize also for the pronunciation Ranyan Timusina Ranyan Timusina Ranyan Timusina Ranyan Timusina I see that they’re not here Lilia Simonian okay not seeing her in the room Andrea Mihalovic president of the world she’s probably busy Lorena

thank you very much I will focus on AI and gender equality in this short intervention AI has enormous potential to advance gender equality but at the same time as we know AI systems can also reproduce existing biases so for instance in the recent study of 2025 AI models were given fictional male and female CVs with the same education experience and job role and they advised the woman to ask for a substantially lower salary than the man and AI -driven bias can also reinforce power imbalances that underpin violence against women. For instance, AI systems may amplify misogynistic content or enable the spread and generation of so -called deepfakes. How do we address such issues? Anticipatory governance plays a key role in ensuring that risks are identified before AI systems are deployed and not only after harm has occurred.

This includes human rights impact assessments that specifically examine risks to equality and non -discrimination, including gender equality. Equal access to technology is also part of the solution because if women are excluded from AI tools and our digital technologies, they are also less likely to be reflected in the data, the assigned choices, and systems that shape future outcomes. And civic participation is very important because people affected by AI systems must be able to understand questions, and help shape how these systems are. governed and used. So this requires awareness raising and capacity building. These elements are reflected in the Council of Europe recommendation on equality and artificial intelligence, which was adopted on the 4th of March, 2026, by the Committee of Ministers.

And on the same day, the Committee of Ministers also adopted the recommendation on accountability for technology -facilitated violence against women and girls, which is not limited to AI, but also calls on member states to address risks arising from AI, including algorithmic amplification of misogynistic content. Thank you very much.

Thank you very much, Lorena, for also tying this conversation to gender equality and civic participation. Arnott, is he in the room? Okay, moving on. Mariam, okay. Yeah, there you go, Mariam.

Greetings once again and apologies for my voice. Clearly yesterday was a success in every dimension. My intervention today addresses exclusion and underrepresentation of youth from marginalized communities in the process of digital transformation. Youth are often addressed as one humongous group based solely on our age. Therefore, whenever we get the opportunity to participate in high -level discussions such as YouthDig and EuroDig, we feel even greater responsibility to represent youth as a whole. But the reality is that we cannot fully do that. I realized this clearly while working on issues of inclusivity in digital messaging and public services. I do not feel equipped to fully represent experiences of young people with disabilities and their needs regarding disability.

I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination.

I am not a person who is a victim of discrimination. I am not a person who is a victim of discrimination. migrant youth, which, given my background as an IDP, remained at the forefront of my mind. Therefore, my sincere request is for the members of these communities to be included in the creation of such platforms and AI systems from the ground up, from the very beginning, started with focus groups, consultations, testing, and continuous feedback mechanisms throughout the entire process of digital transformation of public services. We must ensure that these digital services, including utilized AI -based systems, are designed to suit the people and not the other way around. This topic also raises a question. Considering our growing dependence on the Internet across every aspect of our lives, including access access to public services, when should the access to the Internet begin to be framed as an independent human right within anticipatory governance strategies, sorry, rather than merely a tool of accessing them?

Thank you.

thank you Mariam for contributing to the discussion even though yesterday was a big success it seems Samridhi okay go further

thank you so much my name is Samridhi Rawat I’m a youth digger I want to take this opportunity to speak about something that I think sits at the heart of this question we are discussing here anticipatory governance sounds like a very forward -looking ambition but for millions of people interacting with AI driven public services today the risk of bias exclusion and unequal access is not hypothetical they’re happening right now in real time as we sit here today as a student in the field of informatics leveraging AI for social good and working with data and machine learning models every day I see bias as not a bug that appears by accident but a structural outcome of whose data is used to train these people and systems whose outcomes are used to measure success and whose complaints are visible enough to trigger that correction within the system.

When public services deploy AI without diverse training data, without explainability requirements, which is supposed to be done with proper impact assessment and without accessible feedback mechanisms, they are not making neutral technical choices at all. They are making discriminatory ones, which will affect the whole generations to come. I have had the privilege of not just living in different countries, but in different continents. I have observed two broad philosophies emerging. One prioritizes scale and speed. The other is prioritizing rights and regulation. Both have produced real value and both have produced real blind spots. Speed without accountability embeds inequality in the public system from the very beginning. Caution without inclusion leaves the most marginalized voices out of the design process entirely.

what anticipatory governance must mean in practice is this fairness audits transparency mechanisms and human oversight cannot be afterthoughts added once a system has already gone live and millions of people are using it they must be built in from the very first business line of design and civic participation must go beyond token consultation and i think the youth should be a part of all this conversation thank you

yeah so thank you chair uh a useful start to the discussion on the development of a new system the first point in my opinion is to approach ai and public services as a governance question that happens to involve technology rather than the reverse. So AI Act provides a strong foundation. Now, practical implementation is the hardest, slowest task. So AI governance would not be a conversation for the AI technology per se, and it’s not exhausted by traditional bias and fairness use cases that often frame public discussion. So those use cases matter, but treating AI governance as a discipline in its own right, distinct from the technology and from the normal biased agenda tends to produce better outcomes.

So there are three short roles for this question. One, anticipatory governance means asking the difficult questions before deployment, not after. So who may be excluded, what data is missing, and probably if you don’t have it mentioned in your data, this group will be excluded. You may… person, and kids, which will be more important. Then the civic participation is most useful when it’s structured rather symbolically. So the groups most exposed to exclusion and unequal access are often the first to notice where a system is failing. And last point, AI itself can usually support risk detection, monitoring service uptake across groups, flagging anomalies, surfacing early signals of unequal access, and is a diagnostic instrument, not adjudicator. So final judgment remains with humans and institutions.

Thank you very

Thank you very much, Denys. I see that Jeremy is not online, so we move on to Nadia Simeon. I see that she’s not in the room. Yes, the floor is yours.

Thank you very much for the invitation. I am a deputy director of Ukrainian research company InfoSapiens, and I will tell you briefly about the Ukrainian situation. The full -scale Russian invasion pushed digitalization of public services in Ukraine. And, for example, last year we were rate 40th in the government AI readiness index. In a year we have climbed about 14 steps, and more than 99 % of government services are digitalized. It’s a rather rare situation even for European Union. And 88 % of the population use DIA, which is a system to access information, and we have a very good data on the data. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public.

And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public. And the most important thing is that the information is not available to the public.

And the most important thing is that the information is not available to the public. because of the displacement and other reasons. And so this digitalization, it was really a solution for many people, and it increased access to public services. But still there is age and gender inequality, which is interconnected because women have longer life expectancy. And so there are much more older women than men. So we have conducted in April this year a survey for Council of Europe, and we have seen that due to almost the same usage of different tools of men and women, women have much lower digital literacy than men, especially older women. Women are older than 45 years. I

‘m sorry, I have to interrupt you as we’re very much late on time, and we have to move on to our last guiding question. Yes, I see that there was a hand raised up Yes, I’m sorry, I was out Actually, I was in the list I’m so sorry, but I think we would have to move along because we have roughly 6 -7 minutes until the session concludes and we have another guiding question in the messages So, apologies for this I’ll give the floor back to Gabby So, let’s move to the third question which is the most socially sensible is how can current regulatory approaches be adapted to address real -world implementation challenges especially for vulnerable or underrepresented groups So, let’s hope everybody fits in time and let’s start from Kumhur Er Kumhur, are you here?

No? Elanai Elanai was in another session, I remember Elanai Elanai No Michelle, Valerie, Nikki, Demi no okay parvin parvin jim should do jim

should parvin no good we are saving time brahim brahim time is yours and floor is yours can you hear me uh so i’m brahim you think 2026 and town councillor from monte carlo in foglia italy i would like to thank youth league nadia joao somalia zapata and stephanie stephanie for giving us the opportunity to be here i’ll try to be bring here the voice of the rural community i represent in our country 60 percent of municipalities are composed of less than 5 000 inhabitants we are a majority here there we are a majority all around europe but no one seems to hear us we are affected by pollution created by factories and data centers we are you without gaining any benefit from them we witness the effect of this digitalization and spectators of a process we will never truly be a part of According to the Rural Learning in Digital Index, less than half of households in sparsely populated areas have at least basic digital skills.

Only 20 % have above digital skills and 2 .5 % are ICT specialists. Our schools need improvements in connection and tools. Digital literacy programs need to be implemented in schools and universities, especially when AI is coming out as a key actor in shaping public opinion. Local authorities like ours need funding and guidance on how to implement newer technologies within their workflow. Our community has the potential to become drivers of sustainable change. But if there is anyone with even a single bit of decision -making power at European level listening to me, please believe in me. Please join us and please believe that a new future is possible. Thank you very much.

Thank you, Brahim, for such engaging intervention and for your energy you bring to the floor. Lilith, yes please.

Hello, thank you very much for giving a floor to talk about this very important issue. I’m representing research community from Armenia. And this question is something which I was thinking about last two days, at least during the conference. And I think, actually, we have to go after big players of the market, actually. And I really think that governments need to approach the issue as a real product. Like we have licensing for, or I don’t know, ISO standards for different products in the market. We better regulate, in this sense, also AI and all the products regarding that. Because this is a thing which… enters into our lives and it has to be standardized. It has to be done according to some criteria.

And in that sense, vulnerable and underrepresented groups will for sure take into account and discrimination will be less there, to my mind. And as a researcher, as a representative of a research community, I really think that here we need to do different types of research. Also the perception research, understanding how people overall perceive AI and how they are thinking about, in a philosophical way even, about the AI. Overall, in sociological research, it is very important to come to the definition of an AI. What is it? How is it participating in our society overall? I really want to bring the conversation. And also to this philosophical part of the thematics. And in that sense… we better go deeper into understanding the issue.

Thank you, Lilith, for this philosophical approach, which is also important. And we have another presenter online. Do we have? No. And the last one is Tess. Tess. Oh, there. But you also. Tess.

Thank you. Hello, I’m Tess. I’m part of the Youth League. I’m going to be quick. Just don’t forget transgender identities. For example, in the AI Act, acknowledge gender -based discrimination, but don’t take into account for inclusive gender identities. I’m thinking about the transgender, the non -binary, intersex, and non -gender -confirming people. And this invisibility. I think it’s a great, um, a normative view that trans people do not exist as a population with needs. And trustworthily, AI frameworks focus on making more transparency, but ignore the political assumptions baked into the training of data, the categories themselves. AI bias is not a technical issue, it’s a deeply entrenched social and legal challenge. The norms baked into tech are neutral, they are like political decisions made by specific people, and regulations need to address that, not just make the existing norm more efficient, or the risk will be to reinforce rather than mitigating existing inequalities.

So look at us and include us. Thank you.

Thank you, Tess. And we have one more intervention from the audience. if you could fit in one minute

of course thank you very much my name is federica nori i’m a member of the italian parliament a special representative on ai at the oc parliamentary assembly so we just spoke about ai as a tool of democratic governance and public trust however and if my question is directed towards the minister of justice of georgia however we know that ai can also become a tool of control georgia is a case in point for over 550 days peaceful protesters have faced chinese -made cameras capable of facial recognition emotion analysis and real -time biometric identification used to identify individuals and issue finds how do you think and what’s your opinion on the balance that we should find in order to use ai as a service for people and not to control them as is the case for massive biometric recognition thank you

thank you very much for the question george If I caught your name correctly, as we’re over time and we have to wrap up to messages, I would invite you to discuss with Dimitri, as I assume he will be around for 15 minutes more, the answer to this question. And, okay, I think, okay, sorry, we have to move on, I’ve been told. Then I’d like to give the floor to Milica to wrap up our messages for this session.

Hello, everyone. Thank you for a very valuable contribution by all speakers who are present here and also online, and also everyone who contributed to this afternoon discussion. We have tried to reflect point by point several very important points, and we gave them the title, but I think throughout the cleaning of the document, which remains after, titles will be removed. They are here just to pass the essence of the message. So, let’s go for the first one. Trustworthy AI is a public good. In the public sector, trust is the foundation on which the effective institutions and meaningful public services are built. Trustworthy AI is therefore not only about safe technology. It is about safeguarding democratic legitimacy, human agency, inclusion, and public trust.

AI should be treated as a critical societal infrastructure alongside healthcare, education, welfare, and civic communication. I’m calling if anyone has anything against or have very strong objections. This is like in the wedding. If you have only very, very strong objection against, then say it. If not, you can remain silent. It’s not about semantics. It’s about just the essence of the core message. And we want to be coherent in this conclusion that everybody agrees with the main what was said. So no objections, right? For now, let’s move forward. Okay, so we’re happy. Didn’t misinterpret. Second message is equality, human rights, and standard -based governance. Equality bodies and human rights institutions are essential to addressing algorithmic discrimination in public, public administration’s use of AI.

especially where information and power asymmetries affect individuals’ ability to challenge harm. Are we in consensus on this message, or anyone has a strong objection? Let’s move to the third. Okay, very good. Human -centered public services. Efficiency cannot be only a measure of success in the public sector. AI should improve services not only from the time and cost efficiency perspective, but also make them fairer, more accessible, transparent, inclusive, and trustworthy through the human -centered design focused on citizens’ rights and needs. It seems that a lot of this was put in from the audience on this aspect. So everybody agrees on this human -centric approach. Everybody wants to be centric in the decision -making. And fourth, okay, so meaningful oversight and public accountability.

Human oversight must be real, not symbolic. Public authorities need the capacity to understand, question, override, and remain accountable for AI -supported decision. Trustworthy AI is achieved not only through regulation, but also when fairness, accessibility, inclusion, and accountabilities are built in by design and experienced by all citizens in practice. What about this? All good? Thank you. All good for humans. Okay. Very keen to continue and lead us to the next stage. Human rights -based risk governance. Human rights -based frameworks provide a foundation for trustworthy AI, ensuring alignment with democracy and rule of law. Risk -based approaches support practical tools for risk analysis, stakeholders’ engagement, and mitigation of bias, exclusion, unequal access, and impacts on the vulnerable groups.

Any objections? There’s no same. And the last one. Okay. Technical standards, skills, and global cooperation. Trustworthy AI requires strong governance, technical standards, interoperability, and digital skills. With hundreds of AI -related standards already developed globally, the challenge is to translate this expertise into practical implementation of inclusive policies. Capacity -building efforts are therefore essential to help countries assess their own readiness, absorb global expertise, and advance responsible human -centered digital transformation across the sectors and borders. Thank you. I

It’s very well -formulated messages, and congratulations, Milica, on capturing all what has been said in this audience by the interventions and the speakers. So, can we agree on these messages, which could be polished, and if you have some remarks on what to, you know, polish on the semantics, you can actually contribute maybe by writing to organizational group. But for today, we will… I want to thank, on behalf of all panel and of me and my co -moderator, Aisha, for your contributions, for your active participation, and especially for youth, which takes such a big voice in saying what they want and how they see the future. So I think this is a very commendable experience and what you are bringing to EURODIG Aisha, maybe you would like to say something.

Yes, also I wanted to echo Gavia’s points. If you have any minor corrections or additions you would like to make to the messages, of course, you can reach out to us. And thank you very much for your active participation, and we will not take more of your time from the break. Thank you. And our contributor to final messages, Milica, from Council of Europe, wants to contribute.

Yes, I want to just thank to Valentina Sandic, who basically worked with me hand in hand. It is from ITU, so they are very joint and very synergetic collaboration here. Thank you so much.

Thank you very much. This was a very lovely time. without my mic it’s not going to work thank you very much gavia thank you all for joining us so we now have our final coffee break and back here at four o ‘clock thank