What Proliferation of Artificial Intelligence Means for Information Integrity?

Summary

This discussion, hosted by Latvia’s UN mission in Geneva, focused on the implications of artificial intelligence for information integrity in our global information environment. The panel brought together experts from the UN Human Rights Office, academia, fact-checking organizations, and civil society to examine how AI is transforming how we create, consume, and verify information.

The speakers identified several key risks that AI poses to information integrity. These include the proliferation of deepfakes and synthetic content, AI-generated hallucinations that spread false information, and the potential for discriminatory content moderation systems. Particularly concerning is the use of AI by malicious state actors to manipulate information, conduct surveillance, and suppress dissent, with Russia’s disinformation campaigns regarding Ukraine cited as a stark example. The panelists noted that while individual deepfakes haven’t yet caused measurable behavior changes, the mere existence of AI-generated content is creating widespread skepticism and eroding trust in institutions and information sources generally.

However, the discussion also highlighted AI’s potential benefits for information integrity. AI tools can help fact-checkers work more efficiently, enable better detection of disinformation campaigns, and provide new ways to educate different audience segments about media literacy. The speakers emphasized that different demographic groups require tailored approaches to information literacy, from Ukrainian refugees needing trustworthy local information to elderly populations requiring specialized digital education programs.

The panel concluded that addressing AI’s impact on information integrity requires coordinated action across multiple stakeholders, including governments, tech companies, academia, and civil society, with particular emphasis on transparency, investment in trust and safety measures, and maintaining human intelligence alongside artificial intelligence capabilities.

Keypoints

## Overall Purpose/Goal

This was a panel discussion organized by Latvia at the WSIS Plus 20 review event, focusing on the implications of artificial intelligence for information integrity. The goal was to examine how AI is changing the global information environment, assess the risks it presents to democratic societies, and explore potential solutions that governments, UN institutions, civil society, and academia can implement.

## Major Discussion Points

– **AI’s Transformative Impact on Information Environment**: Speakers emphasized that AI is rapidly changing how information is created, distributed, and consumed, with technologies like ChatGPT reaching 800 million weekly users. However, the full implications are not yet understood, and the pace of change is outstripping our ability to comprehend and respond effectively.

– **Emerging Risks and Threats**: Key concerns include AI-generated disinformation, deepfakes, content hallucination, malicious use by state actors (particularly Russia’s information warfare against Ukraine), micro-targeting of vulnerable populations, and the potential for AI content moderation to perpetuate discrimination and bias.

– **Information Nihilism and Declining Trust**: A critical finding that audiences are becoming increasingly skeptical of all information due to AI’s existence, leading to a “nothing is real” mentality. This erosion of trust in institutions and information sources poses significant challenges for democratic discourse and decision-making.

– **Need for Targeted, Evidence-Based Solutions**: Speakers stressed the importance of understanding different audiences (elderly, rural populations, refugees, youth) and developing tailored approaches rather than one-size-fits-all solutions. They emphasized the need for AI literacy combined with critical thinking skills, and warned against binary approaches to content moderation.

– **Multi-Stakeholder Response Strategy**: Discussion of roles for various actors including: academia maintaining strong human intelligence and knowledge dissemination; fact-checkers adapting tools and methods; governments developing appropriate regulation without censorship; and civil society organizations continuing transparency and accountability efforts despite reduced funding and political pressure.

## Overall Tone

The discussion maintained a consistently serious and somewhat pessimistic tone throughout. Speakers repeatedly acknowledged the gravity of the challenges, with phrases like “not a great landscape,” “very challenging,” and warnings about “blindly driving into a fog.” While some positive applications of AI were mentioned (such as human rights monitoring and fact-checking tools), the risks and challenges clearly dominated the conversation. The tone remained professional and academic, with speakers demonstrating deep expertise while expressing genuine concern about the rapid pace of change and society’s ability to adapt appropriately.

Speakers

– **Zaneta Ozolina** – Professor at University of Latvia, leads a project using AI to address disinformation, expert on information manipulation and young audiences

– **Ivars Pundurs** – Latvian ambassador to Geneva for United Nations institutions

– **Graham Brookie** – Vice President and Senior Director at the Atlantic Council, leads technology programs, expert on information manipulation, builds the Digital Forensic Research Lab

– **Viktors Makarovs** – Special Envoy on Digital Affairs at the Latvian Ministry of Foreign Affairs, moderator of the discussion

– **Peggy Hicks** – Director of the Thematic Engagement, Special Procedures and Rights to Development Division at the UN Human Rights Office

– **Septiaji Nugroho** – Co-founder and chairman of Mafinda (fact-checking organization in Indonesia), fact-checking expert

– **Audience** – Various audience members including Ila from CDAC Network, Claudio (high school student from Romania), and Boris Engelsson (freelance journalist)

**Additional speakers:**

– **Martin Stateris** – Colleague mentioned by Ambassador Pundurs as event organizer, ending his posting in Geneva

# Artificial Intelligence and Information Integrity: A Multi-Stakeholder Discussion on Global Challenges and Solutions

## Executive Summary

This panel discussion, hosted by Latvia’s UN mission in Geneva as part of the WSIS Plus 20 review event, brought together leading experts to examine artificial intelligence’s impact on information integrity. The discussion featured Peggy Hicks, Director of the Thematic Engagement, Special Procedures and Rights to Development Division at the UN Human Rights Office; Professor Zaneta Ozolina from the University of Latvia; Graham Brookie, Vice President and Senior Director of the Atlantic Council’s Digital Forensic Research Lab; and Septiaji Nugroho from Mafindo, Indonesia’s fact-checking organization.

The panel identified critical challenges including AI-generated deepfakes, synthetic content, and discriminatory content moderation systems. Particularly concerning was the documented use of AI by malicious state actors, with Russia’s disinformation campaigns regarding Ukraine serving as a contemporary example. However, the most significant finding was that while individual deepfakes have not yet caused measurable behavioral changes, the mere existence of AI-generated content is creating widespread skepticism and “information nihilism” where “everything is possible and thus nothing is real.”

Despite these challenges, speakers highlighted AI’s potential benefits for information integrity, including enhanced fact-checking capabilities, better detection of disinformation campaigns, and innovative approaches to media literacy education. The panel emphasized that different demographic groups require tailored approaches, from Ukrainian refugees needing trustworthy local information to elderly populations requiring specialized digital education programs.

The discussion concluded with consensus that addressing AI’s impact requires coordinated action across governments, technology companies, academia, and civil society, with key recommendations including increased transparency in AI development, substantial investment in trust and safety measures, and maintaining human intelligence capabilities alongside artificial intelligence systems.

## Opening Context and Framework

Ambassador Ivars Pundurs opened by noting this represents the third consecutive year Latvia has convened this discussion, demonstrating sustained commitment to addressing AI’s impact on information environments. He emphasized how state actors are using AI to manipulate information and conduct surveillance, specifically citing Russia’s AI-driven narratives about Ukraine.

Moderator Viktors Makarovs, Special Envoy on Digital Affairs at the Latvian Ministry of Foreign Affairs, framed the discussion with Yoshua Bengio’s metaphor: “We’re blindly driving into a fog, and one of the areas where this seems to be true is the impact of AI on our information world, on the epistemology of the world.” This captured the fundamental uncertainty about AI’s impact on how we understand and process knowledge itself.

## AI’s Transformative Impact on Information Environment

### Rapid Technological Change and Scale

Speakers agreed that AI is rapidly transforming information environments in ways not yet fully understood. Graham Brookie noted that ChatGPT has reached massive scale adoption, demonstrating AI’s rapid integration into information systems. Septiaji Nugroho observed that people are applying more “gas” than “brakes” to AI adoption, with AI accelerating content creation and dissemination across all sectors.

Peggy Hicks emphasized that AI is changing information environments in ways we don’t fully comprehend, with platforms potentially fragmenting and evolving beyond their current forms. This transformation extends beyond technical capabilities to affect the very epistemology of information—how we determine what is true.

## Emerging Risks and Threats

### AI-Generated Disinformation and Synthetic Content

Peggy Hicks outlined key concerns including AI-generated hallucinations spreading false information, deepfakes convincingly impersonating real people, and biased content moderation systems perpetuating discrimination. These risks are compounded by questionable data provenance, making it increasingly difficult to trace information origins and reliability.

Professor Zaneta Ozolina emphasized that AI enables more sophisticated disinformation campaigns with well-planned tactics, moving beyond random narratives to strategic information operations. This represents a qualitative shift from opportunistic false information to systematic, AI-enhanced manipulation campaigns.

### Malicious Use by State Actors

Graham Brookie noted that AI is being rapidly adopted by bad actors, particularly state actors, for coding narratives and understanding cultural context in information operations. This represents a significant escalation in capabilities available to those seeking to manipulate information environments, enabling more sophisticated targeting and more convincing content creation.

### Micro-Targeting Vulnerable Populations

Septiaji Nugroho highlighted AI’s ability to enable micro-targeting of specific audiences such as elderly people and migrant workers for scams and manipulation. This capability allows malicious actors to craft highly personalized disinformation exploiting specific vulnerabilities, fears, and cultural contexts of particular demographic groups.

## The Information Nihilism Challenge

Perhaps the most significant finding was what Graham Brookie termed “information nihilism”—where “everything is possible and thus nothing is real.” Public polling data showed audiences becoming generally more skeptical and less trusting of institutions, creating an environment where AI’s mere existence undermines confidence in all information sources.

Brookie noted: “We’re seeing trust go down just because AI exists and thus people are a little bit more skeptical of navigating online information environments.” This represents a profound shift where the potential for AI manipulation creates doubt about all information, regardless of actual source or veracity.

Interestingly, Brookie reported that “there hasn’t been one case that we have seen in any place around the world where something like a deepfake or a single piece of synthetic content led to immediate behavior change.” The closest example was Slovakia’s parliamentary elections, but even there, direct behavioral impact was unclear. This suggests the immediate threat may be less about specific synthetic content causing direct harm and more about cumulative effects on trust and information processing.

## Opportunities and Positive Applications

### Enhanced Fact-Checking and Verification

Despite significant risks, speakers identified ways AI can positively contribute to information integrity. Septiaji Nugroho highlighted that AI chatbots can make fact-checking databases more accessible to the public, citing recent implementations improving public access to verified information. He also mentioned Google’s Synth ID technology as a promising development for content authentication.

Peggy Hicks noted that AI tools can help understand global situations in deeper, more nuanced ways at scale and in real-time, potentially improving human rights monitoring and documentation.

### Educational and Outreach Applications

Professor Ozolina emphasized that AI can assist in developing educational curricula and information packages for critical thinking, particularly for reaching underserved social groups such as rural populations. Through her AI for Debunk project, she noted that Ukrainian refugees need trustworthy European information rather than generic anti-disinformation messaging, highlighting the importance of context-specific approaches.

## Addressing Different Audiences and Digital Divides

### Tailored Approaches for Diverse Demographics

A key theme was recognizing that different audiences require specially designed approaches. Professor Ozolina noted that special programs are needed for digitally less educated elderly populations, often implemented through libraries at local levels. She referenced Latvia’s “Seniors Digital Year” program as an example of targeted intervention.

Septiaji Nugroho emphasized that different approaches are required for elderly audiences compared to young people, such as specialized digital academies. Both speakers highlighted particular challenges faced by elderly and rural populations in navigating AI-enhanced information environments.

### The Role of Traditional Institutions

An important theme was the critical role of traditional institutions like libraries in addressing modern AI challenges. Speakers noted that libraries and librarians possess traditional skills in information integrity that remain highly relevant in the digital age, serving as trusted local resources for digital literacy education.

## Educational Approaches and Critical Literacy

Septiaji Nugroho made a crucial distinction, arguing that “AI literacy should be accompanied with AI critical literacy, just like when we do critical thinking on digital literacy.” He emphasized the importance of Socratic prompting techniques and noted that information consumption patterns need to change from vertical to lateral reading when comparing AI-generated and traditional sources.

Professor Ozolina provided crucial framing: “It’s not about debating pro or con. Artificial intelligence is here to stay. So therefore, the question is how to balance human intelligence and artificial intelligence.” This moved the discussion beyond resistance versus acceptance to focus on practical coexistence strategies.

## Institutional Responses and Challenges

### Inadequate Current Responses

Speakers expressed concern about current institutional responses. Peggy Hicks noted that “government responses tend toward binary solutions that don’t work from a free expression standpoint and can enable censorship of dissent.” Graham Brookie complemented this by noting “there’s large-scale retrenchment from industry transparency efforts and reduced investment in trust and safety fields.”

### The Need for Evidence-Based Approaches

Peggy Hicks emphasized the importance of evidence-based approaches that identify good practices and incorporate academic research. She noted the need to “address these issues on a firm information basis… But also at pace,” highlighting the tension between evidence-based policy-making and urgency created by rapid technological change.

## Stakeholder Roles and Responsibilities

### Technology Platforms and Corporate Responsibility

Septiaji Nugroho argued that platforms should bear primary responsibility for identifying synthetic content through proper watermarking, noting that “fact-checkers now face impossible demands to verify whether content is synthetic, which should be platforms’ responsibility through proper watermarking.”

### Academic and Civil Society Roles

Professor Ozolina outlined academia’s role in “spreading knowledge, communicating with different societal groups, and filling information vacuums.” However, this was challenged by audience member Boris Engelsson, a freelance journalist, who questioned academic credibility, citing concerns that “most medical research in the past 50 years may be fake.”

Graham Brookie emphasized civil society’s continued importance in driving transparency and accountability, despite facing “challenges and reduced funding and political pressure.”

## Audience Engagement and Additional Perspectives

The discussion included several audience interventions that enriched the conversation. A Romanian high school student named Claudio asked about recent presidential elections and disinformation campaigns. Audience member Ila from the CDAC Network raised questions about blockchain technology for enhancing information integrity and transparency.

Questions were also raised about the role of libraries and librarians in information integrity, reflecting broader interest in how traditional information institutions can adapt to AI challenges.

## UN Human Rights Office Initiatives

Peggy Hicks outlined specific initiatives the UN Human Rights Office plans to implement, including developing a Human Rights Digital Advisory Service referenced in the Global Digital Compact to help states and businesses navigate AI challenges. The office also plans to continue the BTEC project encouraging companies to describe their human rights practices and promote best practices.

## Unresolved Questions and Future Challenges

The discussion revealed significant gaps in understanding AI’s actual impact on information environments. Graham Brookie emphasized the need to “collect more case studies and data to have higher confidence assessments about AI’s impact on information environments.”

Peggy Hicks raised fundamental questions about platform sustainability, noting that “platforms may not exist in their current form in the near future,” and asked “what type of narrative response is useful in countering disinformation without amplifying it?”

## Conclusion

The discussion demonstrated remarkable consensus on the fundamental nature of challenges while acknowledging significant disagreements about specific solutions. The conversation highlighted that while AI presents serious risks to information integrity, it also offers significant opportunities for enhancing our ability to create, verify, and disseminate reliable information.

The key challenge lies not in choosing between human and artificial intelligence, but in developing approaches that effectively combine both while maintaining critical thinking capabilities and institutional safeguards that democratic societies require. Different communities, cultures, and contexts require tailored approaches that build upon existing strengths and address specific vulnerabilities.

The path forward requires continued dialogue, research, and experimentation, combined with commitment to evidence-based approaches that can evolve as understanding of AI’s impact deepens. Most crucially, it requires maintaining focus on preserving and strengthening democratic discourse and decision-making processes as technological foundations of information systems continue to evolve rapidly.

As the moderator noted in closing, while we may be “blindly driving into a fog,” the collective expertise, commitment, and collaborative spirit evident in this conversation provide grounds for cautious optimism that these challenges can be navigated while preserving the values and capabilities that democratic societies require to thrive in an AI-enhanced world.

Ivars Pundurs: THE remainder of the episode is about the collapse and recession of the IMF by IWM. It’s basically about our relationship with the older generation and politics all over the world, and the implications that some things clarify. I’ve talked to people who have interested in my channel because I want to share my psychology there with them. Recording in progress. Hello, good afternoon. I am a Latvian ambassador here in Geneva to the all possible United Nations institutions and it falls on me to make an opening speech of this what I hope will be very interesting and stimulating discussion. So I’ll just get on with it. So excellencies, ladies, gentlemen, dear friends and colleagues, I’m delighted to welcome you to this panel discussion on the implications of artificial intelligence for information integrity. Like most countries, Latvia has entered the AI race. We have established a national AI center that brings together the public and private sectors, as well as academia to foster rapid AI adoption. But speed alone is This is the third consecutive year that Latvia has convened a discussion on how artificial intelligence affects our information environment. While technology has advanced a lot in this time, the issue at hand has only grown more urgent. This discussion is particularly relevant in the context of the ongoing WSIS Plus 20 review. When the WSIS action lines were first formulated over two decades ago, few could have anticipated the AI revolution. And yet many of those action lines on access to information and knowledge, the ethical dimensions of the information society, trust and security in the use of ICTs, and the role of media remain highly relevant. At the same time, the transformative change we have witnessed in recent years makes clear that information integrity has emerged as a distinct and critical challenge requiring focused attention and collective action. Threats to information integrity and to our society stem not only from the technology itself, but also from its malicious use by state actors. These actors harness artificial intelligence to manipulate information, shape minds, and behavior, conduct surveillance, censor, and suppress dissent. Such practices undermine democracy, erode societal cohesion, and jeopardize international peace, as well as our shared efforts to achieve the sustainable development goals. In Europe, the stark example is Russia’s use of AI-driven tools to spread narratives aimed at justifying its unprovoked war of aggression against Ukraine, a war that flagrantly violates international law, inflicts immense human suffering, and devastates infrastructure. Given the scale of these risks, it is essential that we engage in open dialogue, share experiences, and explore practical solutions. This panel provides an important opportunity to do so. do just that. I extend my sincere appreciation to our distinguished speakers from international organizations, civil society and academia who will share their insights today. I also thank each of you for taking your time from your busy schedules to join this discussion. I would like to express my gratitude also to the organizers of the WSIS plus 20 high-level event for making this session possible. I wish us all a thought-provoking and productive exchange. Last but not least, I wish to thank my dear colleague Martin Stateris who has put this event together. It is the last event for him in his posting in Geneva, so I wish all the best to him in his future career. And now it is my pleasure to give the floor to Mr. Viktors Makarovs, Special Envoy on Digital Affairs at the Latvian Ministry of Foreign Affairs who will moderate today’s discussion. Thank you.

Viktors Makarovs: Thank you very much, Ambassador, for the introduction. To introduce the topic again very quickly, I think the important thing to say is that it’s a juncture of two themes that obviously are on many people’s minds. One is AI. I don’t have to go into that because we are at an AI event, basically. But the other one is information integrity, which is a very important though recent and not quite well-known idea of an information environment that is global, open, free, but at the same time safe and secure. And we will be looking at how these two important phenomena that we are dealing with today are playing together and what they mean. Another thing to know for you about this discussion is that this is not the first of its kind that we have organized. We organized a similar event last year, and indeed Peggy Hicks, one of our speakers, was kind enough to participate in that conversation as well, and we have done it on other platforms and in other forums over the last two or three years. The reason why we want to repeat this exercise is that the world we live in has changed dramatically. Technologically speaking, in terms of AI, we’re in a completely different place. The technology has evolved, but also adoption has evolved or increased exponentially over the last year. Just ChatGPT alone has 800 million weekly users today, and it’s about four times increased since last year when we had the first installment of this discussion. One of the people looking closely at the risks presented by AI, Yoshua Dengio, said that we’re blindly driving into a fog, and one of the areas where this seems to be true is the impact of AI on our information world, on the epistemology of the world, and we want to address this and find out what the state of play is, but also what the outlook on AI and the risks it presents to information environments is today, and then the next stage will be to talk about things to do. What can governments do? What can the United Nations do? Civil society and academia, and to do that, we have four speakers, three here with us, and one will be joining us online. I will introduce them as they are invited to speak. We will have very quick interactions because we’re really short on time, and hopefully, After these interactions, we’ll have time to engage the audience as well. So as the first speaker, I would like to give the floor to Peggy Hicks, who is Director of the Thematic Engagement, Special Procedures and Rights to Development Division at the UN Human Rights Office. And the question I have for you, Peggy, as well as for the other speakers in this first round is, how is AI changing our information environment? What risks does it present to information integrity today and also in the near future? And can it also be a technological force for good and help support information integrity, open, free and trustworthy information? So over to you.

Peggy Hicks: Viktors, you always ask the easy questions. There is a lot tucked in there So I would say three pieces. The first is really looking at how the environment is changing. And I think it’s really important to say that this is a rapid change in the information environment that we really do not even yet understand fully where we’re going. I listened to a podcast recently where they were talking about the fact that social media may ultimately be turning into the much more fragmented platform-based, protocol-based approach that, you know, some of us had wondered if it might deliver something different and better than the existing platforms. But it’s coming at a time where there’s actually a question whether platforms themselves will continue to exist in the way that they do. So I do think in this information integrity realm, one of the greatest things I want to emphasize is that we need to address these issues on a firm information basis. And that’s why I’m glad we have academia with us and work being done in those areas. But also at pace. We can’t address the problems of yesterday rather than today. And I’m afraid we are too often in danger of doing that. Secondly, looking at what are the risks that we see now. And it’s really, it’s not a great landscape, but there is to say, there is so much that’s happening and we hear a lot about dis- and misinformation. Obviously, those problems exist regardless of generative AI and large language models, but let’s throw in. Developing the whole ideas of hallucination, which doesn’t seem to be getting any better. The possibility that we all rely on information that’s not what we hope it to be is very, very real, and then, of course, we have the issues around deepfakes and the impact there. Those are two of the areas that are most often talked about, but I also think we have to talk about the fact that AI content moderation that is likely to happen may be infused with some of the same flaws we see within AI machine learning generally, where it could fuel discrimination or exacerbate some of the problems that we already see, and, of course, this is all a landscape that’s based on data that we’re not really sure of the provenance or use of that data and the privacy of it as well, so there are all sorts of risks that we’re facing in terms of the AI itself, but the second set of risks I think we have to emphasize is how will government respond to these issues because, unfortunately, there, too, the landscape is not very promising. What we’ve seen is a tendency to look for sort of binary solutions. Information is good or information is bad. It’s true or it’s false. We can flip a switch and solve the issue of disinformation if only the companies were willing to do it. That is not going to work from a free expression standpoint, from a fact that there are always going to be facts that are contested and that there will always be actors who want to use that lever to flip the switch against to censor speech or to censor dissent in ways, so we have to be very careful about the tools that we deploy to address these problems as well, but finally, you asked me, is there anything good here? Do you have anything positive on the landscape for us to look at, and I think from a human rights perspective, we see lots of value here. We do understand that these tools are tools that will help us to understand what’s happening in the world in a much deeper and more nuanced way at scale and in real time in a way that could allow us to better address human rights problems globally. One of the things that we have always focused on in the human rights movement is that when there’s a spotlight on a situation, it tends to be much harder for things to happen in a negative way with regards to human rights. The fact that we now have access to data sets that we never did before and the fact that we’ll be able to allow people in general to engage in the human rights cause in new ways I think is very promising, but it’s making sure that those positive aspects keep up with some of the risks that we see that’s the real challenge. Thanks.

Viktors Makarovs: Thank you very much, Peggy. Looks like the risks part of the equation at the moment is much more weighty than the other way around. Professor Zaneta Ozolina, at University of Latvia, you do lots of things and the teams you work with, you are engaged, and I think the University of Latvia actually leads a project that is aimed at using AI to address disinformation, but I know you also look closely at the audiences, especially young people who are the potential targets and victims of information manipulation. How does it look from your perspective?

Zaneta Ozolina: Thank you. Thank you for inviting and to share some thoughts. So actually, I was prepared, Viktor, to respond to the questions which you already asked to Peggy, but I will leave them for the later stage. So, but I will respond directly to your question. And I think that it is very important whenever we are addressing information integrity or disinformation issues to know exactly what audience we are addressing and which audience we are targeting. And information integrity as such is a very relevant issue, but different audiences have completely different attitudes, what is relevant or not. For instance, when we started our project, AI for Debunk, then we found out different target audiences which are targeted by disinformation. And we also decided to interview Ukrainian refugees, expecting a priori, let’s say our hypothesis was that this is one of those target groups, which is very much exposed to Russia’s disinformation. And what we found out that this is a completely different situation, because mostly these are women who arrived with their children as refugees running away from war. And they know what war is about and what Russia is up to. They don’t have to be convinced about the disinformation’s role in their lives. But what they were lacking, they were lacking information about what is happening in Europe, in language which they could understand, information sources they could trust. So, therefore, it is a different issue which we are traditionally used to discuss in our countries. And it also highlighted how information integrity is relevant for different target groups, but in a very specially designed way and very specially prepared narratives and messages. Another point which we addressed in our project was comparative analysis of two big cases, War in Ukraine and Climate Change, and it was a targeted selection of those two cases because it’s interesting to see what are the differences and what are the similarities which ERAO actually highlights, which are disinformation campaigns, whether these are just random disinformation narratives thrown in different media sources or it is a well-planned strategic campaign. And what we found out is that even narratives could be different, the tactics, how it is applied, the models which are used, they are part of very well-planned campaigns. Another point, so you should control my time, so another part of the project is that we work together very closely with different IT companies because one task of the project is to elaborate a special tool which could assist in identifying disinformation. Honestly, I was quite skeptical at the very beginning, and honestly I’m sharing, this is my attitude, because so I was always questioning how artificial intelligence can actually compete with human intelligence in identifying disinformation campaigns, but in the end I found that there will be very excellent extra products produced in the end, because it’s not only about one universal tool which could work in all situations and in all target groups, but this tool will assist in, for instance, developing new curricula for schools. These tools will assist to produce some special information packages to be used for developing critical thinking again in different societal groups. It will assist to address Those for instance, social groups and societies that are not very keen on using classical information sources. Like, for instance, people in rural areas do not read the Financial Times and the Washington Post. But they also need information on what is happening in the political landscape.

Viktors Makarovs: Thank you very much. Actually, you partially answered the second question I had in mind, but that’s excellent. So, we’re very practical here looking at things to do. Now, the next speaker I want to go to is joining us remotely, Septiaji Nugroho, who is co-founder and chairman of Mafinda, which is a fact-checking organization in Indonesia, and I hope we have him online. Yes, I’m here. Fantastic, Septiaji, nice to hear you again. So, the first question I would like you to address is like, great to see you as well, is the same I asked the other speakers, which is about the state of play on AI and information integrity, where we are today, compared to maybe a year ago, when you also joined our discussion here at WSIS, where this is heading. Please, three, max four minutes. Thank you.

Septiaji Nugroho: Thank you, Mr. Ambassador, and also Latvia Mission for UN for inviting me. So, basically, even before the effectively work on combating those video disinformation. And now we arrive in AI era disinformations that actually AI bring both immense opportunities but also face significant challenges. So the AI itself, because they can accelerate content creation and disseminations, we feel that there are people now put more gas on AI using compared to the breaks that we want to have. So there are several risks that we are already facing. For example, now people are asking to us whether this content is synthetic or not synthetic, which is actually shouldn’t be our problem. It should be the job of especially digital platform to make sure that people understand whether this content is synthetic or not synthetic. Because if it is now delegated to the fact checkers, no fact checkers in the world can face this, so many AI contents already. And this deepfakes and synthetics media, they are now being really, really realistic. And the problem is that not every platform provide enough watermark that can be detected. Maybe a platform like Google, they put like Synth ID, then they also introduce the detector, so we can get help to detect very accurately whether the content is synthetic or not. also have now in a very difficult position. The tech platform also now, because of what happened in the United States, is also affecting how fact-checkers are now operating. So AI problems definitely is going to be a very, very big problem for fact-checkers around the world. Septiaji, just one additional question. Do you also register in your work user of AI to micro-target audiences in a way that perhaps is not really visible to the others? Does it show up as an issue? Well, yes, definitely. I think AI, because AI can also micro-targeting, especially now with what we see is using AI for scams, for example. So they can also target specific people because they have already database, they can target like elderly people, they target like migrant workers. using very convincing videos and audio. I think this is also like a big problem that now we are facing. Although in other side, we also try to make use of AI. For example, now we just launched like two weeks ago using chatbot AI. So to make sure that people can connect to our database well, because before using AI, it is quite difficult for people to try to use our database. But now it is going to be much, much easier. But I feel that the challenge is still much, much bigger than the opportunities that we can have.

Viktors Makarovs: Septiaji, thank you very much. We will come back to you in short order. So, fourth speaker, last but not the least, Graham Brookey, you are Vice President and Senior Director at the Atlantic Council. You lead the technology programs and it is a sprawling business. But we know that you are an expert on information manipulation, been in this area for many years, fighting it, obviously. So, what is your very concise take on the AI landscape today and tomorrow?

Graham Brookie: So, my job at the Atlantic Council, first and foremost, thank you. My job at the Atlantic Council over the last eight years is to build a thing called the Digital Forensic Research Lab, which is a team of open-source researchers spread out across 17 different countries on four different continents. And so, while my accent would indicate that I am from a US-based headquartered organization, our work is very much global. Now, on the question of how AI is changing information integrity, I think that there’s, number one, a scoping question. Number two, some key findings on narratives themselves. Number three, on how the tech is changing. And then, number four, on how stakeholders are changing. And so, the scoping question, how AI is changing the information environment, we’re not just talking about mis- and disinformation. We’re not just talking about what state actors are doing in the information environment. We’re talking about a really, really broad set of potential online harms, including basic scam activity, which was just mentioned, including things like CSAM, including any number of things that happen in online spaces that have negative externalities for society. And so, while we tend to talk about mis- and disinformation, most of all in this space, it is part of a much larger ecosystem of problems that we’re trying, or opportunities, in some cases, that we’re trying to meet. Now, what we’re seeing in terms of how AI is changing the information landscape from our research, I think there’s four key findings. And I will use a case study from the United States. I would be very, very clear in the sense of the United States is a very large media market where a lot of these platforms happen to be headquartered, but they aren’t that immune. I’m going to start by saying that the AI information ecosystem isn’t immune to vulnerabilities. In fact, it’s rife with vulnerabilities. And so I say that with a little bit of humility. The first finding is that on AI’s impact in the information environment, it’s early days. How bad actors are navigating or harnessing AI to do bad activity in an information environment or manipulative activity in an information environment, we don’t have that much amount of data yet. And I say that as the pace of technological change is very, very rapid. And so to Peggy’s point, we’re talking about platforms that might not be platforms here in the very near future. And yet our data sets are still early days. We need to collect more information. We need to collect more case studies in order to have higher confidence assessments. That’s number one. Number two is in everything that we are monitoring, we are seeing higher amounts of content that is created by generative artificial intelligence, GAI. So we are seeing in any given case, whether that’s scam activity, whether that’s election integrity or election information environments around election processes, we are seeing more content that is GAI. Now the third finding is that that doesn’t necessarily equate to behavior change. There hasn’t been one case that we have seen in any place around the world where something like a deepfake or a single piece of synthetic content led to immediate behavior change. We’ve seen some cases where it’s gotten close, very, very close, but for the most part, institutions around information integrity have been able to quickly identify and create conversation around that. The highest and best example on a deepfake changing the result of the election being a very near case in Slovenia, parliamentary elections is the main case study that has been mentioned most of all in Slovakia, sorry, Slovakia parliamentary elections. That’s my mistake entirely. There hasn’t been a single instance where a deepfake has had an immediate measurable change on behavior. In addition to that, we would generally agree with the assessment from some of the large platforms like OpenAI, where the GAI content that we are seeing doesn’t necessarily lead to more engagement on other platforms. To make that more granular, if a The piece of content is synthetic. If it’s a deepfake or a G.A.I. image or synthetic audio and then it spreads on other social media platforms, the fact that it was created by artificial intelligence doesn’t necessarily mean that audiences are engaging with it at a higher pace. And then the fourth thing is a very pessimistic view. What we are seeing from public polling is that audiences are generally more skeptical and less trusting of institutions in general and the information environment specifically. And so we’re in this situation that is accelerated or informed by this moment of A.I. in which everything is possible and thus nothing is real. Take that from a number of other foreign policy examples. And that’s not great for democratic outcomes. It’s not great for multi-stakeholder outcomes. It’s not great for the integrity of the information environment at large. So we’re seeing trust go down just because A.I. exists and thus people are a little bit more skeptical of navigating online information environments. The ecosystem changes. We’re seeing large-scale retrenchment from industry and transparency efforts. We’re seeing less investment in fields of trust and safety from governmental institutions, from civil society institutions. And so the landscape is very challenging. And then from a technical standpoint, I think it remains to be seen whether the actual tech change of A.I. is changing or increasing the defender’s dividend. And so whether A.I. tooling specifically for trust and safety is actually having a net positive impact. What we are seeing is pretty rapid adaptation by bad actors using A.I. And so where, for instance, state actors, we are seeing more use of A.I. by state actors in information operations in particular for things like coding narratives, for things like understanding cultural context or breaking down language barriers. And so we’re seeing rapid adaptation there. And it remains to be seen whether the defender’s dividend ticks up over time.

Viktors Makarovs: Okay, thank you, Graham. That was not a very rosy outlook from you either. Now, we don’t have as much time to address the second and probably first important question, what do we do with it? So I would ask you to perhaps offer a really brief take, if you can do it within two minutes, that would be fantastic. Go in the same order, Peggy. So, you know, you work with the focus on human rights. So a very important part of the UN, what can your system, so to say, do to help countries and civil societies address this issue? And what are the most international processes that we also as member states and stakeholders should be paying attention to? Again, easy questions.

Peggy Hicks: Yes, in short order. No, I mean, I think there is a lot that’s happening to address these issues. So, you know, I can’t give a full overview of it now. But I mean, I think one of the key things that we need to do, and actually A.I. can help us do this, is we need to look for good practices where things are being handled in a way that is human rights respecting and does a better job and incorporates the evidence and data. I mean, that’s the real problem here is that we tend to think we know what the answers to questions are without actually having looked at what are the real problems and what works to address them. So bringing in the academic community more successfully and allowing the good practices to be paralleled in a variety of different geographic and resource environments is crucial. We are trying to set up something called the Human Rights Digital Advisory Service as referenced in the Global Digital Compact that we hope will help us to do that and be a real resource with an academic network behind it of helping states and businesses to navigate this space. The second piece of it that we’re engaged in is that there’s a lot of talk, and I think it will be a big part of the conversations in this setting in general around WSIS and Tech for Good, is around the role of the companies. And the reality is we need to find a way that encourages, incentivizes and holds accountable companies for the way that they’re engaging in this space. We have a project called BTEC, which is a way to get companies to describe what are their human rights related practices for us to pull out the good examples from that and to try to push a race to the top and in the same time to distinguish between companies that are making the right type.

Viktors Makarovs: Fantastic. And you mentioned academic input. So Professor Rosalina, what is Latvian academia doing together with civil society? What’s going to be your input to address the issue?

Zaneta Ozolina: Yeah, indeed. Again, one of the questions which need definitely more time, I would divide my answer in two groups. So one would be more general character. When it comes to what we could do, I think it’s very important to balance our attitude towards artificial intelligence. Now it seems that society is divided into groups. Those who are praising results of artificial opportunities of artificial intelligence and those who are denying. So it’s not about debating pro or con. Artificial intelligence is here to stay. So therefore, the question is how to balance human intelligence and artificial intelligence. The other point is I also would like to join Peggy. It’s very important to think about new ways how to regulate and also how to govern artificial intelligence and the way how it’s influenced and how it leaves its impact on information integrity. As far as academia is concerned, I think one of very important points is to keep human intelligence strong and powerful is just to spread knowledge and to contribute to knowledge. There is no other way how to stay sober in this very digitalized world is to be very well equipped with knowledge. So that’s what academia can do. The second point is that academia knows how to communicate with society, how to communicate with different societal groups and faster artificial intelligence and actually technologies are growing. More communicators and mediators will be needed. And we are here to communicate with the younger generation, to communicate with those who are in need. And the third point which I wanted to mention is that today’s discussion is about information integrity. It’s very important to avoid information vacuum because what we are very often observing in the public space, in education system, in the way how younger generation receives and consumes information, it has a lot in common with vacuum. So therefore it’s important for politicians, for academia representatives to be more and more investing in information integrity which replaces information vacuum.

Viktors Makarovs: Thank you. Keeping human intelligence strong, that’s quite a challenge. We’ll try to do it. We go to Septiaji Nugroho now, if you’re still with us, a quick sort of advice from you. As a leading fact-checker, how are you adapting to the AI age and what are the lessons and your advice for other fact-checkers and perhaps those who support you? Again, please very quickly, just two minutes if you may. Thank you.

Septiaji Nugroho: Yeah, basically Mofindo is working. Hello, good evening. As I said, we are showing in two areas. One is fact-checking. Of course, AI is going to one way to help the fact-checkers to pinpoint very quickly the information that we are doing, the fact-checking. And also we are also using AI to connect our staff to personalize our content to be easier or digital literacy educations. Now government also speeding up on doing AI literacy, but we feel that sometimes they work in the wrong directions because they forget that AI literacy should be accompanied with AI critical literacy, just like when we do critical thinking on digital literacy. That’s why our role at the moment is helping, assisting when the government now initiate the curriculum for coding and programming and also AI as early as elementary schools. We are now involved in making sure that AI critical literacy is a big part of that to make sure that they are not only learning about prompting, especially for senior high schools, but also how to critically, how to understand how to use like a prompting in Socratic prompting to make sure that people shouldn’t especially don’t lose their ability to critical thinking on AI literacy. Thank you.

Viktors Makarovs: Thank you very much, Septelji. And Graham, same question by and large goes to you. What can we expect from civil society organizations like yours to do to address AI and how can other stakeholders support you?

Graham Brookie: You can expect us to remain engaged, which is maybe not a particularly novel thing to say, but it is increasingly difficult in this landscape. Right now, there is a narrative, including from some partner governments, that any effort to regulate this space or to moderate this space or to create transparency in this space equates to censorship. And that is fundamentally not true. And so we have to address that narrative. And number two, the investment in this landscape is retrenched and we have to drive more investment into these sets of issues in order to have. The Basic Transparency that Allows the Space for Accountability and Rights-Respecting and Rights-Protecting Approaches. And here, the prioritization has got to be on transparency for frontier models for this AI moment. Number two, trust and safety, an investment in trust and safety, especially by industry and especially as it relates to generative artificial intelligence’s impact on the information environment. Number three, most broadly, investment in the critical institutions that create and protect and sustain the multi-stakeholder system that has kept things like the Internet open, secure, and interoperable, the things that WSIS absolutely stands for. And so that’s absolutely critical. And I say investment in the institutions like academic institutions that are driving long-form research, like fact-checking organizations or independent media that create open dialogue, and like civil society organizations that are driving technical research.

Viktors Makarovs: Oh, fantastic. Thank you very much. Now, we have about five minutes to address questions. And I have to mention that because we are online as well, there have been some interesting comments like one member of the audience writing that we should also recognize the role of libraries and information services and the traditional skills of librarians. There is also a question about the use of blockchain technology to enhance integrity of information, trust, and transparency. Not exactly our topic there, perhaps, but also a question if information integrity is essential to mitigate the bad or adverse effects of disinformation and disinformation. I think it’s obvious. So we’ll perhaps leave it out there for the time being. Let’s take questions from the audience. Finally, I can see one hand there and you two. Let’s start with you, please. And please introduce yourself. And please, really, a short question.

Audience: Yes. So my name is Ila. I think you need to speak into the mic, so there’s really… My name is Ila, I’m with the CDAC Network, Communicating with Disaster-Affected Communities. And on your point about information nihilism, as it may or may not relate to Gen-AI, studies show that people’s willingness to believe false or true information is not really connected to the level… of Realism of the Content, but rather factors like repetition, narrative appeal, perceived authority, etc., etc., and the viewer’s state of mind. A key element of a long-term, portfolioed approach to respond to that would be to fund and support local independent journalism, but realistically funding prospects are bleak for local journalism. So how could Gen AI be used to provide information that meets those epistemic and social and psychological needs, say, to help local human rights-based CSOs create counter-narratives to incendiary information? Or should we avoid that, completely using Gen AI to create counter-narratives?

Viktors Makarovs: Okay, that’s one question out there, I think this goes to… Let’s take one or two more. I can see the hand over there, please.

Audience: Can you hear me? So hi, my name is Claudio. And I’m a high school student from Romania. I wanted to ask you this question because in our very recent presidential elections we actually had a huge disinformation campaign, I think you may have heard about it. And in your discourse from earlier, I heard a lot about educating the new generation. But what do you do with the people that remain illiterate digitally? The people from the countryside, the older people, don’t you think there is a need for regulation on the AI companies that generate the watermark or something like that? Thank you. Very good question. I would say two of them. But let’s take another one, please. Yes, Boris Engelsson, a journalist freelance. I heard praise about the academia to be able to safeguard some information integrity. One year back at the University of Geneva, medical faculty, there was a big symposium with a big shot of this domain claiming that most of medical research in in the past 50 years is fake. And they quoted the former editor-in-chief of The Lancet, I think, who confirmed that. I would have more questions, but I will stick to that one. So I’m not quite sure I understood your specific question, sir, to this panel. The question is that if even nearly medical faculties, like faculty, I am not even considered economics and psychology, which has been disqualified as science since long. But I mean, if even medical researchers confess now that they consider most of medical research fake, are these people the best source of integrity of information because they are called academics? Oh, that seems to be an interesting one to address.

Viktors Makarovs: Sorry. Sorry. Yes. We have a question on regulation of AI again, use of AI to actually inform and create counter-narratives. And also, what do you do with audiences that lack the AI literacy and AI critical literacy, as we just heard from one of the speakers? Who would like to address perhaps very quickly each of the panelists?

Peggy Hicks: Sure. I think each of the questions has its answer built in it to some extent. I think we have to explore. We’re actually doing some of that work internally and happy to talk more about it. But what type of narrative response is useful? But that’s part of what I meant. And I have to say, I think you’re absolutely right. There’s probably research out there that’s not good. But like with anything, we can’t just say there isn’t any good research. We have to actually look at the research, see how it’s done, and vet it to ensure it’s solid. But one of the things that we’ve learned from some of the research on disinformation is that some of the things that we think would be useful to counter it don’t work, and the alternative is true. That we need to make sure that we’re not actually bolstering disinformation by giving it greater breadth by responding to it in the wrong way. And so I really like the idea of how do we bring those pieces together and engage more. On your question, I think it’s a really valid one that making sure that our education systems, both for younger and older people, are helping people.

Zaneta Ozolina: There is no universal remedy to all those questions which were raised, because very often disinformation and also information integrity is country-specific. And regarding your question, a very similar problem we have in Latvia. Also, the elderly generation digitally is not so educated as the younger generation. And this is a government policy at the moment that this year is called Seniors Digital Year, when special programmes are offered on the local level, particularly to senior groups, in order to be prepared for the next rounds of digitalisation. And this, by the way, to a very large extent, is executed by librarians and in the libraries. So therefore, the question which was raised before, what is the role of libraries? Very great role, particularly addressing those target groups which are remotely placed.

Viktors Makarovs: Thank you. Septiaji, if you’re still there, we’ll get back to you for your quick comments. Let’s do it right now. Yes, 60 seconds max, please.

Septiaji Nugroho: Yes, I think for the first one, it is going to be very urgent that we need to explain that how we read usually information vertically needs to be changed to be read laterally. So that’s why we want to make sure that people need to exercise their freedom of expression. the way they are reading information, especially comparing analytically information coming from NGINI and also with the information from the library and everything. And the second one about how we target the elderly. Mafindo is now running the third year as we have an elderly digital academy. So basically we have a specific approach which is different when we target the young audience. So that’s why we would really like to share the experience. We are very happy to share you if you want to connect us on how the Indonesian version of elderly digital academy work, especially also on AI. You can contact me. Thank you.

Viktors Makarovs: Thank you very much, Septiaji. Something very specific there. And Graham, 30 seconds. And then we are finished and we have to leave.

Graham Brookie: I’m happy to get more in depth on the very complex conversations after the session. The advice is very simple. Yes, we have to avoid the oxygen of amplification, but you have got to engage better across all that. Don’t be afraid of your own shadow. Engage in the information environment that you have, not the information environment that you want. Thank you very much.

Viktors Makarovs: And the only thing I have to say, two things. Thank you very much to the speakers, to the audience, of course, also those online. And the second thing I have to say is there is coffee courtesy of this particular side event. So you’re welcome to enjoy it in the break. Thank you very much. And have a nice day.

Agreement points

AI is rapidly transforming the information environment in ways that are not fully understood

Speakers

– Peggy Hicks
– Viktors Makarovs
– Graham Brookie

Arguments

AI is rapidly changing the information environment in ways we don’t fully understand yet, with platforms potentially fragmenting and evolving beyond current forms

We are blindly driving into a fog regarding AI’s impact on our information world and epistemology

While AI-generated content is increasing across all monitored cases, there hasn’t been documented behavior change from single deepfakes or synthetic content

Summary

All speakers acknowledge that AI is fundamentally changing how information is created, distributed, and consumed, but emphasize that we lack complete understanding of these changes and their implications

Topics

Content policy | Human rights principles | Interdisciplinary approaches

AI enables sophisticated targeting and manipulation by malicious actors

Speakers

– Ivars Pundurs
– Zaneta Ozolina
– Septiaji Nugroho
– Graham Brookie

Arguments

State actors are using AI to manipulate information and conduct surveillance, with Russia’s AI-driven narratives about Ukraine as a stark example

AI enables more sophisticated disinformation campaigns with well-planned tactics and models, not just random narratives

AI enables micro-targeting of specific audiences like elderly people and migrant workers for scams and manipulation

AI is being rapidly adopted by bad actors, particularly state actors, for coding narratives and understanding cultural context in information operations

Summary

Speakers agree that AI significantly enhances the capabilities of malicious actors to create targeted, sophisticated information manipulation campaigns

Topics

Cyberconflict and warfare | Content policy | Human rights principles

Different audiences require tailored approaches to information integrity and AI literacy

Speakers

– Zaneta Ozolina
– Septiaji Nugroho
– Peggy Hicks

Arguments

Different audiences require specially designed approaches to information integrity, as demonstrated by Ukrainian refugees needing trustworthy European information rather than anti-disinformation messaging

Different approaches are required for elderly audiences compared to young people, such as specialized digital academies

Need for evidence-based approaches that identify good practices and incorporate academic research rather than assuming solutions

Summary

Speakers emphasize that effective responses to AI and information integrity challenges must be customized for specific demographic groups and their unique needs

Topics

Online education | Digital access | Human rights principles

AI presents both significant risks and potential benefits for information integrity

Speakers

– Peggy Hicks
– Zaneta Ozolina
– Septiaji Nugroho

Arguments

AI tools can help understand global situations in deeper, more nuanced ways at scale and in real-time, potentially improving human rights monitoring

AI can assist in developing educational curricula, information packages for critical thinking, and reaching underserved social groups like rural populations

AI chatbots can make fact-checking databases more accessible to the public, as demonstrated by recent implementations

Summary

While acknowledging serious risks, speakers agree that AI can be leveraged positively for education, accessibility, and human rights monitoring when properly implemented

Topics

Online education | Digital access | Human rights principles

Similar viewpoints

Both speakers express concern about inadequate institutional responses to AI challenges, with governments seeking overly simplistic solutions and industry reducing investment in safety measures

Speakers

– Peggy Hicks
– Graham Brookie

Arguments

Government responses tend toward binary solutions that don’t work from a free expression standpoint and can enable censorship of dissent

There’s large-scale retrenchment from industry transparency efforts and reduced investment in trust and safety fields

Topics

Legal and regulatory | Human rights principles | Freedom of expression

Both speakers advocate for balanced, critical approaches to AI education that maintain human cognitive abilities while leveraging AI benefits

Speakers

– Zaneta Ozolina
– Septiaji Nugroho

Arguments

Importance of balancing human intelligence with artificial intelligence rather than taking pro or con positions

AI literacy education must be accompanied by AI critical literacy, including Socratic prompting techniques to maintain critical thinking abilities

Topics

Online education | Critical thinking | Interdisciplinary approaches

Multiple speakers recognize that elderly and digitally illiterate populations require special attention and targeted interventions to address AI-related information challenges

Speakers

– Zaneta Ozolina
– Septiaji Nugroho
– Audience

Arguments

Special programs are needed for digitally less educated elderly populations, often implemented through libraries at the local level

Different approaches are required for elderly audiences compared to young people, such as specialized digital academies

There is a need for regulation requiring AI companies to implement watermarking, especially for digitally illiterate populations

Topics

Digital access | Online education | Legal and regulatory

Unexpected consensus

The limited immediate behavioral impact of AI-generated content despite increased production

Speakers

– Graham Brookie
– Peggy Hicks

Arguments

While AI-generated content is increasing across all monitored cases, there hasn’t been documented behavior change from single deepfakes or synthetic content

AI presents risks through hallucination, deepfakes, biased content moderation, and questionable data provenance that threaten information reliability

Explanation

Despite widespread concern about AI-generated content, there’s surprising agreement that single pieces of synthetic content haven’t yet caused measurable behavior change, suggesting the threat may be more about cumulative effects and trust erosion rather than immediate manipulation

Topics

Content policy | Human rights principles | Cyberconflict and warfare

The critical role of traditional institutions like libraries in addressing AI challenges

Speakers

– Zaneta Ozolina
– Audience

Arguments

Special programs are needed for digitally less educated elderly populations, often implemented through libraries at the local level

Libraries and librarians should be recognized for their traditional skills in information integrity and their role in supporting digital literacy programs

Explanation

There’s unexpected consensus that traditional information institutions like libraries are crucial for addressing modern AI challenges, highlighting how established information literacy skills remain relevant in the digital age

Topics

Online education | Digital access | Capacity development

Overall assessment

Summary

Speakers demonstrate strong consensus on the fundamental challenges posed by AI to information integrity, the need for tailored educational approaches, and the dual nature of AI as both threat and opportunity. There’s also agreement on the inadequacy of current institutional responses.

Consensus level

High level of consensus on problem identification and general solution directions, with speakers complementing rather than contradicting each other. This suggests a mature understanding of the issues and potential for coordinated policy responses, though implementation details may require further discussion.

Different viewpoints

Role of platforms vs. fact-checkers in synthetic content identification

Speakers

– Septiaji Nugroho
– Peggy Hicks

Arguments

Fact-checkers now face impossible demands to verify whether content is synthetic, which should be platforms’ responsibility through proper watermarking

Need for evidence-based approaches that identify good practices and incorporate academic research rather than assuming solutions

Summary

Nugroho argues that platforms should bear responsibility for identifying synthetic content through watermarking, while Hicks emphasizes the need for evidence-based approaches and good practices rather than assuming technological solutions will work

Topics

Content policy | Liability of intermediaries | Digital standards

Trust in academic institutions as guardians of information integrity

Speakers

– Zaneta Ozolina
– Audience

Arguments

Academia must focus on spreading knowledge, communicating with different societal groups, and filling information vacuums

Academic credibility in information integrity is questionable given admissions that most medical research in the past 50 years may be fake

Summary

Ozolina advocates for academia’s central role in maintaining information integrity through knowledge dissemination, while an audience member challenges academic credibility by citing concerns about falsified research

Topics

Interdisciplinary approaches | Human rights principles | Content policy

Unexpected differences

Effectiveness of AI tools in combating disinformation

Speakers

– Zaneta Ozolina
– Septiaji Nugroho

Arguments

AI can assist in developing educational curricula, information packages for critical thinking, and reaching underserved social groups like rural populations

AI chatbots can make fact-checking databases more accessible to the public, as demonstrated by recent implementations

Explanation

While both speakers acknowledge AI’s potential benefits, Ozolina was initially skeptical about AI competing with human intelligence in identifying disinformation but became convinced of its utility for educational tools. Nugroho, while implementing AI solutions, emphasizes that challenges are ‘much, much bigger than the opportunities.’ This represents an unexpected nuanced disagreement about AI’s net benefit despite both using it practically

Topics

Content policy | Online education | Interdisciplinary approaches

Overall assessment

Summary

The discussion revealed relatively low levels of fundamental disagreement among speakers, with most conflicts centered on implementation approaches rather than core principles. Key areas of disagreement included the division of responsibility between platforms and fact-checkers, the role of academic institutions in information integrity, and the relative balance of AI’s benefits versus risks.

Disagreement level

Moderate disagreement with significant implications – while speakers generally agreed on the problems and broad solution categories (education, regulation, transparency), their different approaches to implementation could lead to conflicting policy recommendations. The disagreement about academic credibility is particularly significant as it challenges a foundational assumption about expertise and authority in information verification.

Partial agreements

Similar viewpoints

Both speakers express concern about inadequate institutional responses to AI challenges, with governments seeking overly simplistic solutions and industry reducing investment in safety measures

Speakers

– Peggy Hicks
– Graham Brookie

Arguments

Government responses tend toward binary solutions that don’t work from a free expression standpoint and can enable censorship of dissent

There’s large-scale retrenchment from industry transparency efforts and reduced investment in trust and safety fields

Topics

Legal and regulatory | Human rights principles | Freedom of expression

Both speakers advocate for balanced, critical approaches to AI education that maintain human cognitive abilities while leveraging AI benefits

Speakers

– Zaneta Ozolina
– Septiaji Nugroho

Arguments

Importance of balancing human intelligence with artificial intelligence rather than taking pro or con positions

AI literacy education must be accompanied by AI critical literacy, including Socratic prompting techniques to maintain critical thinking abilities

Topics

Online education | Critical thinking | Interdisciplinary approaches

Multiple speakers recognize that elderly and digitally illiterate populations require special attention and targeted interventions to address AI-related information challenges

Speakers

– Zaneta Ozolina
– Septiaji Nugroho
– Audience

Arguments

Special programs are needed for digitally less educated elderly populations, often implemented through libraries at the local level

Different approaches are required for elderly audiences compared to young people, such as specialized digital academies

There is a need for regulation requiring AI companies to implement watermarking, especially for digitally illiterate populations

Topics

Digital access | Online education | Legal and regulatory

Key takeaways

AI is rapidly transforming the information environment in ways not yet fully understood, creating both opportunities and significant risks to information integrity

While AI-generated content is increasing across all monitored cases, there hasn’t been documented behavior change from single deepfakes or synthetic content pieces

Different audiences require specially designed approaches to information integrity – there is no universal solution

AI literacy education must be accompanied by AI critical literacy to maintain human critical thinking abilities

The challenge requires balancing human intelligence with artificial intelligence rather than taking binary pro/con positions

Evidence-based approaches incorporating academic research are essential rather than assuming solutions

Public trust in institutions and information sources is declining, creating an ‘everything is possible, nothing is real’ environment

Investment in trust and safety, transparency for frontier AI models, and support for critical institutions is crucial

Resolutions and action items

UN Human Rights Office to develop a Human Rights Digital Advisory Service as referenced in the Global Digital Compact to help states and businesses navigate AI challenges

Continue the BTEC project to encourage companies to describe their human rights practices and promote best practices

Academia to focus on spreading knowledge, communicating with different societal groups, and filling information vacuums

Fact-checking organizations to adapt by using AI tools to connect databases with users and develop educational content

Civil society organizations to remain engaged despite challenges and drive investment in transparency and trust and safety

Governments to implement specialized programs for digitally less educated populations, particularly elderly groups, often through libraries

Unresolved issues

How to effectively regulate AI companies to ensure proper watermarking and content identification

What to do with digitally illiterate populations, particularly elderly and rural communities

How to use AI to create effective counter-narratives without amplifying disinformation

Whether AI tooling for trust and safety is having a net positive impact (defender’s dividend unclear)

How to address the fundamental challenge that platforms may not exist in their current form in the near future

How to balance free expression concerns with the need to address AI-generated disinformation

How to maintain institutional trust while addressing legitimate concerns about AI-generated content

Suggested compromises

Avoid binary solutions to information problems and instead focus on evidence-based approaches that respect free expression

Combine AI literacy with AI critical literacy education rather than focusing solely on technical skills

Use AI tools to assist rather than replace human judgment in fact-checking and content verification

Engage in the information environment as it exists rather than as we want it to be

Focus on transparency and accountability measures for companies rather than outright restrictions

Develop specialized approaches for different target audiences rather than one-size-fits-all solutions

We’re blindly driving into a fog, and one of the areas where this seems to be true is the impact of AI on our information world, on the epistemology of the world

Speaker

Viktors Makarovs

Reason

This metaphor effectively captures the fundamental uncertainty and philosophical implications of AI’s impact on how we understand and process knowledge itself. It frames the discussion not just as a technical challenge but as an epistemological crisis affecting the very foundations of how we know what we know.

Impact

This framing elevated the discussion from technical concerns to deeper philosophical questions about truth and knowledge, setting the stage for speakers to address both immediate risks and fundamental challenges to information integrity.

We can’t address the problems of yesterday rather than today… We need to address these issues on a firm information basis… But also at pace

Speaker

Peggy Hicks

Reason

This highlights a critical paradox in policy-making around rapidly evolving technology – the need for evidence-based responses while moving quickly enough to remain relevant. It challenges the traditional approach of thorough study before action.

Impact

This comment established a tension that ran throughout the discussion between the need for careful research and the urgency of the AI transformation, influencing how other speakers balanced immediate concerns with longer-term solutions.

Everything is possible and thus nothing is real… We’re seeing trust go down just because AI exists and thus people are a little bit more skeptical of navigating online information environments

Speaker

Graham Brookie

Reason

This captures a profound psychological and social consequence of AI – that its mere existence creates a crisis of epistemic confidence even before specific harms occur. It identifies ‘information nihilism’ as perhaps more dangerous than specific disinformation.

Impact

This insight shifted the discussion from focusing solely on technical solutions to addressing the broader erosion of trust in information systems, leading other speakers to consider psychological and social dimensions of the problem.

There hasn’t been one case that we have seen in any place around the world where something like a deepfake or a single piece of synthetic content led to immediate behavior change

Speaker

Graham Brookie

Reason

This empirical observation challenges common assumptions about AI’s immediate impact on behavior, suggesting that fears about deepfakes may be overblown while the real damage is more subtle and systemic.

Impact

This finding provided important nuance to the discussion, helping ground fears in actual evidence and redirecting attention from spectacular individual cases to systemic effects on trust and information processing.

It’s not about debating pro or con. Artificial intelligence is here to stay. So therefore, the question is how to balance human intelligence and artificial intelligence

Speaker

Zaneta Ozolina

Reason

This reframes the entire debate from resistance versus acceptance to integration and balance, moving beyond binary thinking to focus on practical coexistence strategies.

Impact

This perspective helped shift the discussion from defensive measures against AI to proactive strategies for human-AI collaboration, influencing how other speakers approached solutions and adaptation strategies.

AI literacy should be accompanied with AI critical literacy, just like when we do critical thinking on digital literacy

Speaker

Septiaji Nugroho

Reason

This distinguishes between technical AI skills and critical thinking about AI, highlighting that teaching people to use AI tools is insufficient without teaching them to question and evaluate AI outputs.

Impact

This insight influenced the discussion of educational approaches, emphasizing that solutions must go beyond technical training to include critical thinking skills, which other speakers then incorporated into their recommendations.

There is no universal remedy to all those questions which were raised, because very often disinformation and also information integrity is country-specific

Speaker

Zaneta Ozolina

Reason

This challenges the assumption that global problems require universal solutions, emphasizing the importance of local context, culture, and specific vulnerabilities in addressing information integrity.

Impact

This observation helped ground the discussion in practical realities, leading speakers to consider how solutions must be adapted to different contexts rather than seeking one-size-fits-all approaches.

Overall assessment

These key comments collectively transformed what could have been a technical discussion about AI tools into a nuanced exploration of epistemological, psychological, and social challenges. The most impactful insights reframed fundamental assumptions – moving from viewing AI as a problem to be solved to understanding it as a reality requiring adaptation, from focusing on spectacular individual harms to recognizing systemic erosion of trust, and from seeking universal solutions to acknowledging contextual complexity. The discussion evolved from initial concerns about specific AI capabilities to deeper questions about how societies can maintain information integrity while adapting to technological transformation. The interplay between empirical findings (like the lack of documented deepfake behavior change) and philosophical observations (like ‘everything is possible, nothing is real’) created a sophisticated dialogue that balanced immediate practical concerns with longer-term societal implications.

How can AI be used to provide information that meets epistemic and social psychological needs to help local human rights-based CSOs create counter-narratives to incendiary information?

Speaker

Ila (CDAC Network)

Explanation

This addresses the practical application of AI for positive counter-messaging while considering the psychological factors that influence belief in information

Should we avoid using Gen AI completely to create counter-narratives?

Speaker

Ila (CDAC Network)

Explanation

This explores the ethical and practical considerations of using AI-generated content to combat disinformation

What specific regulatory measures should be implemented for AI companies regarding watermarking and content identification?

Speaker

Claudio (high school student from Romania)

Explanation

This addresses the need for technical solutions and regulatory frameworks to help users identify AI-generated content

How do we address digital illiteracy among older populations and rural communities in the context of AI-driven information environments?

Speaker

Claudio (high school student from Romania)

Explanation

This highlights the challenge of protecting vulnerable populations who may lack the skills to navigate AI-enhanced information landscapes

How can blockchain technology be used to enhance integrity of information, trust, and transparency?

Speaker

Online audience member

Explanation

This explores alternative technological solutions for ensuring information authenticity and traceability

What is the role of libraries and information services and traditional skills of librarians in maintaining information integrity?

Speaker

Online audience member

Explanation

This examines how traditional information institutions can contribute to combating AI-driven misinformation

Whether AI tooling specifically for trust and safety is actually having a net positive impact – does it increase the defender’s dividend?

Speaker

Graham Brookie

Explanation

This addresses the effectiveness of AI-based solutions in defending against AI-generated threats and whether defensive capabilities are keeping pace with offensive ones

How do we collect more case studies and data to have higher confidence assessments about AI’s impact on information environments?

Speaker

Graham Brookie

Explanation

This highlights the need for more comprehensive research and data collection to better understand the evolving landscape

What type of narrative response is useful in countering disinformation without amplifying it?

Speaker

Peggy Hicks

Explanation

This explores the strategic communication challenges of responding to misinformation without inadvertently spreading it further

How do we ensure that our responses to AI-driven information threats don’t inadvertently bolster disinformation by giving it greater breadth?

Speaker

Peggy Hicks

Explanation

This addresses the unintended consequences of counter-disinformation efforts and the need for evidence-based approaches

How do we balance human intelligence and artificial intelligence in information environments?

Speaker

Zaneta Ozolina

Explanation

This explores the fundamental question of maintaining human agency and critical thinking in an AI-dominated information landscape

How do we develop new ways to regulate and govern artificial intelligence’s impact on information integrity?

Speaker

Zaneta Ozolina

Explanation

This addresses the need for updated governance frameworks that can effectively manage AI’s impact on information systems