Digitaler Salon: Massenpropaganda auf Knopfdruck

Digital Salon, a series on questions of the networked present from the Alexander von Humboldt Institute for Internet and Society. Hello and a warm welcome to the Digital Salon here at the Alexander von Humboldt Institute for Internet and Society. I am incredibly happy that you are all here today and in such large numbers. That is truly fantastic. I also look forward to you joining the discussion later, and of course, I’m very pleased that so many of you are also joining us via the livestream on Alex Berlin. There are three elections coming up: in Mecklenburg-Western Pomerania, in Saxony-Anhalt, and in Berlin. And the election campaigning has already begun. We have all already noticed it. There are election posters out on the streets . The first televised debates are now following. For example, the first televised debate for Saxony-Anhalt is this evening, right after our Digital Salon, and of course there is street campaigning, but today, a large part of the election campaign also takes place in our feeds on social media, and there’s a new development there. Since October, stricter rules for political advertising have applied in the EU, and Meta and Google have therefore decided to largely discontinue political ads. So, for the most part, they no longer exist there. But I would like to ask now, who here in the room has recently seen political content regarding the election in their feeds anyway? Okay, interesting. Not that many general posts or other forms. Well, I have definitely received a lot regarding the election. Exciting. I think that also has to do with, well, what kind of feed you have and perhaps whether you have a lot of political content in your feed in general or not. But one can say that politics has not disappeared from social media. However, nowadays it depends on organic reach, meaning posts without paid advertising on social media like Instagram and others that are still being displayed. And so, a lot depends on the algorithms , on the content, and also on the users and their preferences regarding what we see in our feeds today. That is why we want to talk today about how all of this is changing online election campaigning. And what happens when generative AI is added to the mix and automated content generated by AI floods our feeds? My name is Felix Leitmeier. I am a journalist, and I am very happy to be able to guide you through the evening today, and above all, I am pleased to have three wonderful guests with me who bring very different perspectives. A big round of applause for the three of them, please. Well, I’ll start with Professor Jeanette Hofmann. You are the founding director of research here at the HIEG and a political scientist. You are a professor emerita for Internet Policy at FU Berlin and a member of the newly created media council for public broadcasting. In addition, your research focuses on disinformation, AI and democracy, and the relationship between digitalization and democracy. It’s great to have you here, although I must say, since this is your home base, I feel more like I’m the one visiting you. A big round of applause, please, for Professor Sonert-Hofmann. Also here is Dr. Eva Flecken. You have been the director of the Media Authority Berlin-Brandenburg since 2021 . You were the chair of the Directors ' Conference of the State Media Authorities in 2024/25, and before that , President at Sky Deutschland. There, you led the areas of regulation, youth protection, and policy for the DACH region. Welcome and a big round of applause for Dr. Eva Flecken. Also here is Leon Wedel. You are a research associate and PhD candidate at the Weizenbaum Institute in Berlin and lead the project "Your Feed, Your Choice." Furthermore, you study, among other things, how algorithms on TikTok curate political content for users in a personalized way. So, we will definitely be talking about that a lot today. And you have a goal: to understand the influence all of this has on the democratic public sphere and voting behavior. Um, exactly. Welcome and a big round of applause for Leon Wedel. Well, I’m looking forward to a great discussion today, and I have a question for all of you to start. We just talked about the fact that Meta and Google have decided to largely discontinue political ads in the EU. What does that change for campaigning online? Perhaps you would like to start ? I have no idea what that changes concretely. It certainly changes what we see online, but whether it actually changes people’s voting behavior, I don’t think we can judge that well, because voting behavior—and I hope we will talk a bit more about this—is very multifactorial. It isn't influenced by one channel, but by many factors, some of which have nothing to do with media consumption at all. Good that you bring up that point. That is definitely a point we will be discussing extensively today. I will come back to you on that question later . Otherwise Eva, what would you say? How does that change the campaign and what does it mean for the campaign that there are now, well, fewer ads on social media in the EU? Well, I wish I could come up with a great rebuttal to Jeanette and sort of add fuel to this discussion, but I actually believe— and I want to agree with her here— that it is still way too early to determine what kind of changes this will actually bring, and indeed, voting behavior has such complex reasons or a voting decision has so many causes, that it will be difficult to pinpoint it in the end and say: "Aha, and because Meta and Google, for example, made this decision, the election results turned out differently." I believe, overall, it is still worth looking at what is actually happening around elections, especially in the run-up to elections, on social media platforms, even beyond the question of paid advertising, yes or no, but I am sure we will touch upon this topic today as well. That will definitely be part of it. Leon, what would you like to say? How will campaigning change? How has it already changed because the ads no longer exist? Um, luckily for me, after having studied TikTok for several years now, the campaign on other platforms will become more like TikTok, because paid advertising, political advertising , has not been allowed there since 2019 . Hey, yeah, exciting. You wanted to say something directly to that, I think . You reacted. I was just thinking, um, I was thinking about those big campaign posters you see on the main streets in Berlin, right? They cost a lot of money, but if you ask the parties what exactly they achieve, they say, we can't afford not to hang them up because the others are doing it too. Those are largely imitation effects that don't pay off in concrete results, but rather have to do with doing what the others are doing. And in that sense, if there is no paid advertising on social media, then that is a form of imitation that no longer takes place. Exciting. Okay. Maybe there are other forms that are taking place now. We will talk about that again, where others will surely get caught up as well. Eva, I have a question. Is it in the spirit of the EU now, can you say, the DSA regulation, that there is now largely no political advertising on the platforms? Was that the intention? Well , I obviously can't look inside the head of Ms. von der Leyen or her entire staff, but I suspect that when they conceived this regulation, they didn't have this as a target. So what did the EU Commission want? The EU Commission wanted to implement what I think is a completely obvious objective, namely that we all know why someone is actually communicating something. That means, um, I want to be informed, for example, was I paid to express this opinion now, or did I receive some other form of compensation? It doesn't just have to be money; it could be a trip, it could be anything at all. And I believe that is an interest we all have, that we all recognize, and which, by the way, we are already familiar with from traditional broadcasting. That's why in the classic TV sector, we already have the distinction: this is the editorial part, handled by a journalist according to journalistic standards, and this is advertising. And advertising always has a commercial intent. You are supposed to buy something. Um, and so it is labeled accordingly. And that is no different with political advertising, purchased political advertising. I believe the Commission's goal was to say that if political influence is taking place through monetization, i.e., purchased reach, then it must be labeled accordingly for the sake of the users. Google and Meta and others decided that it was all too vague for them—the definition of political advertising— that it's "legally uncertain," as they always say, and they pulled the plug, as you mentioned in your introduction. I am quite sure that this was not the goal of the EU Commission. And perhaps you could summarize once more, what was the EU's big goal with this regulation? Transparency, quite simply, transparency, and that is actually in the name of the regulation. Uh, transparency regarding the fact that where money flows, there is clearly a purchased interest, and that must be visible to us all. And the second point was this microtargeting, but that is a special case. Um, that would perhaps go beyond the scope here for the moment, but the EU Commission was concerned, as is often the case, with consumer protection, if you will, in the form of transparency. I understand. Jeanette, can there also be an advantage, um, that there is now no more advertising because political visibility no longer depends on advertising budgets, for example? How do you view this topic? Um, we actually know relatively little about how advertising affects people's behavior. Well, I think as long as election research has existed, it has never been able to agree on whether political advertising actually works or not. And that has a lot to do with the assumptions underlying it. Um, you mentioned earlier that I've been doing a lot of work on disinformation lately. The same thing applies there. The research community itself is divided over whether and how something like that works. And that has something to do with the different views of human nature that underlie it, among other things. So, do you fundamentally consider people to be very easily influenced or not? And people who think , for instance, that we can all think for ourselves—that is, those who assume that—would attribute significantly less influence to political advertising, just as they would to disinformation. And depending on which assumptions you base this on, perhaps also how much optimism you have regarding the autonomy of human thought , perception, and action, you will evaluate such things differently. There are people who are perhaps almost cynical, who would claim that these vast sums that political parties spend on advertising remain relatively ineffective. Interesting. Leon, I have a direct question for you now. You examined and took a close look at what is happening on TikTok. Um, and which parties were particularly visible in the feeds. Um, back then during the 2019 federal election, you said earlier , right, uh, there wasn't any paid political advertising on TikTok, which you looked at closely. Um, and uh, just now at the last federal election, there wasn't any then either. Um, and I would like to ask you, what did you observe back then during the federal election? Um, exactly. We also carried out the project during the 2025 federal election. Your feed, your choice. And just as context, so it's really clear what we can find out with this. In the end, we asked voters: "Hey, can you donate to us the data that TikTok collects about you?" It's a tool, it's a process, and everything is GDPR compliant and anonymized. And people participated, um, over 900 people and voters, and also gave us the information, hey, what did you vote for ? And then in the end, we were able to look at, okay, what kind of content ended up in your feed, and is that perhaps related to your voting behavior or not? Um, and what we were able to observe there, um, I think is very exciting, in the end, the 2025 federal election was characterized above all on TikTok and in the social media discourse by two parties, that was the AfD and that was the Left, or Die Linke . Um, and both were successful in the users' feeds in their own way. Um, ultimately, what you vote for is always what's on top, that's also in the feed. So, we weren't able to observe that any party was somehow overrepresented in the feeds of others, of people who wouldn't vote for them. That was not the case. Um, but we were able to observe that the Left and the AfD were successful in their own different ways. That means the AfD was successful primarily because they produced in bulk . There were massive amounts of videos and in the end, the algorithm works like this: the more you produce, the higher your chance of slipping in somewhere you might not have ended up if you produced less, but of course also more into the feeds of your voters . And the Left, on the other hand, doubtlessly had the luck of creating viral momentum through the firewall speech, um, in the context, um, of Heidi Reichinnek in the Bundestag. This speech developed a viral momentum on the platform, where, um, despite very few videos, the videos gained an incredible amount of reach because they simply resonated. And if you want to put it a bit more bluntly, you could say, okay, one did it through mass, the other through class, and that's it. On mass and class, I want to say something about that directly. The question for me, well, one has heard that often too, right? The young people, they, uh, all tended to vote for the fringes rather than the center. And what is the reason for that? The AfD was so present. Can one seriously imagine that someone, let's say, who holds values close to the Greens, watches five videos with the AfD and then votes AfD? No, right? Absolutely not. Yeah, no, I just want to mention that one has to keep that in mind. My impression is that such videos have at most amplifying effects. Meaning, they might make people who would rather go to the lake on Sunday think, okay, I'll go vote. But they don't choose their pol —they don't change their political preferences. That's why I always think, when you discuss in such a context, you have to be careful not to engage in " blaming the messenger." in the sense that the channel is held responsible for the values that voters ultimately develop over long periods of time. I have a follow-up question: is there any scenario at all in which attitudes, and subsequently voting behavior, change as a result of social media usage? Well, there are studies—I’m certainly no expert on this—so I’d rather address this to you as a question. um, which show that, uh, the same thing over and over—repetition, so to speak—does lead to a certain effect, and that there are also connections between classic, perhaps more obvious campaign messages and videos or content that have nothing to do with elections, but rather with leisure behavior, personal attitudes, or similar things, independent of a party. And, um, would you say there is no possibility at all that the attitudes, values, and ultimately the voting behavior of individual people could change due to social media usage? Maybe, Net, maybe you could start. Well , I would say that what parties achieve with their political communication is, among other things, the ability to justify resentments. So, if someone thinks to themselves, "I've never liked those foreigners in the apartment next door," then a party can develop a language and a narrative that gives them a good reason to say that out loud or something like that, but it won't create that resentment itself. In other words, what parties manage to do is give expression to certain feelings. And the AfD is pretty good at that, providing justifications, so to speak, but they will always just be reacting in an amplifying way to something that is already present in the voters. That’s how I would describe it, and it’s not without influence, but you shouldn't overestimate it either; instead, I think we should always take seriously what drives people in such directions in the first place, and especially with young people: why are they choosing the fringes rather than the center? That is interesting to look at, instead of just bashing the parties . Leon, you’re nodding; do you see it the same way? Yes, by and large, uh, of course. Um, but I would like to add that, at the end of the day, of course, the TikTok feed—or any social media feed, that's how it's designed, that's how it's programmed—is a mirror of the person, of the personality. That is why, of course, it is absolutely correct. As a rule, the attitude comes first, and then comes the content that confirms that attitude. Um, but I also believe that, while never social media alone, in combination with other influences, other developments, experiences, and family and friends, the algorithm can lead to a shift in attitude and, potentially, a change in voting behavior; but it’s a long process and one that cannot be applied to society as a whole. Mr. Leon, we were just getting to your analysis. You were already looking at the state elections before we even get to the impact of it all. What are you observing right now on TikTok regarding the state elections? Um, right now we obviously only have monitoring of what parties post, how much they post, and how many views or likes they generate. And here we see significant differences , depending on which state the parties are from. So, we are looking specifically at, okay, are these politicians and party accounts from Saxony-Anhalt, Mecklenburg-Western Pomerania, or Berlin? Um, and here the impression is relatively clear. I'll start with the less exciting part. Um, Mecklenburg-Western Pomerania looks exactly like the federal election effectively. That means we have many videos from the AfD and actually the CDU too, but they don't play a role in views and likes at all anymore. Um, and the AfD can generate views through volume here, but the likes per video and views per video—that is, the engagement—still actually remains ahead for the Left Party. Um, in Berlin , we have the Greens and the SPD, who are actually leading in all metrics. Um , and in Saxony-Anhalt, we ultimately have the AfD, which stands with a massive lead over everyone else. Meanwhile, accounts from AfD Saxony, also from AfD Saxony-Anhalt, and of course especially Ulrich Siegmund and the politicians there have generated over 30 million views. It's always important to note that these views are global. Anyone can see these videos. That means it’s not just people in Saxony-Anhalt. Um, but that is exactly the exciting part that we hope to find out during the course of the study, because we will then have the data donations and see to what extent these videos actually reached the voters from those states. If you had to summarize what has changed from the federal election to the current state elections, can anything be said about that yet? That at least the AfD accounts from Saxony-Anhalt—to use that casual term again—are apparently no longer producing just quantity, but also quality. So, we’re managing both now, so to speak, and you’re saying during the federal election it was more of a choice. Do I go all in on quantity or do I go all in on the topic of quality, you’re saying? Okay. And perhaps in contrast, we actually have the fact that across Germany, the AfD and the CDU have produced almost exactly the same number of videos so far in the election campaign. Um, and yet the CDU is non-existent in terms of views and likes, which of course also has something to do with the platform, with the users on the platform, and which voter groups are even active there. Janette, quantity and quality, are those the two main strategies on social media for you as well, or is there something else we haven't discussed yet ? I can't say because I don't conduct empirical research on that. But I was just wondering about the data regarding Berlin. Um, the previous assessments are that the Greens and the Left dominate very strongly within the S-Bahn ring, while outside the S-Bahn ring in the southeast, I assume, the AfD is stronger, and otherwise in the southwest perhaps the CDU and the SPD. But that can't be mapped with your data donations, right? For that, we would need to receive a great many, let's put it that way. Of course, to be able to map that reasonably and granularly. So, one can probably assume, based on what you see now, that the data donations come from within the S-Bahn ring. The data currently is not yet based on data donations. That is just the public data from the parties. Therefore but yes, will you be able to see that in the future? Uh, if enough data donations come in now, let's assume a lot of Berliners donate their data there. Exactly. If, in the course of the month , we practically get, let's say, 100 data donations per postal code and everyone participates, then we can indeed check such granular details if necessary. What is the current status on that? How many are there per code, so to speak? Not that many yet. We only started yesterday. Um, and experience shows it will take another week before the donation campaign really takes off. Well, then we are curious to see if enough comes together from Berlin so that we can also answer those specific questions. Interesting. Now another big new topic is being added. We spoke about it earlier, it's not actually that new anymore, but in recent years at least, it was AI, right? That was already a topic around the federal election campaign. Back then, um, the parties experimented with it. There was , for example, a report in NRW about a bot in the CDU called Konrad that assisted in drafting press releases and posts. Other parties have also used AI for research, drafting texts, and so on . And now there are new reports from Correctiv, for instance, that the AfD or its affiliates are using an AI tool to automatically create social media content. And Eva, I’d be really interested to know, does it make a difference to media regulators whether a post was created by AI or if it’s, so to speak, handmade? Are those two different categories for you? No, neither for media regulators, I believe , nor for the rest of us. After all, media regulation is provided by all of us. That is important for me to emphasize. Because, for one thing, we might still be able to tell the difference between AI-generated and non-AI-generated content at the moment —though I’m not entirely sure—but let’s check back next year, and I think we can put that to rest. Maybe there will be technical options, but certainly not our eyes, our ears, or anything else that can determine whether something is AI-generated or not. And that is why I think the reasons are obvious to say that even when it comes to regulatory issues, it makes no difference; what matters is the impact. And you just mentioned the use of AI in election campaigns. We took action last year as the Media Authority Berlin-Brandenburg against the AfD in Brandenburg. Because they created an AI-generated spot for the 2024 state election campaign. Perhaps some of you have seen it. In it, there were, let’s say, very blonde, bright families bathed in sunlight contrasted with threatening, dark, darkly dressed people. And we determined—not just us , but the Commission for the Protection of Minors determined—that this way of editing content against each other is likely to frighten those under 16. So, that is a violation of youth protection and "impairs development," as they say in legal jargon, and although this spot was AI-generated, that was completely irrelevant to us. Well, it also didn't make any difference, so to speak, in the, I'd like to say, argumentation, in how we conducted this case, because we said: We have a video and it makes absolutely no difference if I make it with real actors or if I make it with AI-generated actors or performers. This video is capable of frightening people under 16, and I think that is a contemporary stance. And I believe we should all internalize that and not get involved in such a discussion—I don't even know if this discussion exists— but not get involved in a discussion like: "Yes, AI-generated is less bad." That's nonsense; we won't be able to distinguish any of it in a year anyway. Will it happen so quickly that we won't be able to tell if it's AI-generated or not? Oh, we are already at that point today. Well, I believe with certain videos, with certain photos, and also audio tracks, we have long been at the point today where, um, deepfakes and the imitation of existing people work so well, almost perfectly, that, depending on the case, we don't even notice the difference anymore. Yes, I must confess, I have already seen videos where I definitely had to think about it for a few seconds and then also had to look closely at the credits to see if there was perhaps a hint, when one simply couldn't tell anymore, even now, right? Yes, I'm curious. There is another development, an international disinformation campaign. It's called Matryoshka. Um, it involves, well, there are an insane number of false claims currently circulating on social media, and some are also AI-created. For example, voices of politicians and similar figures are being imitated there. Um, there are really, uh, extreme accusations, it must be said, that are being spread in part on social media. We see, for example, how prevalent posts still are on X, claiming that politicians allegedly committed murders , or that the same politician in this case, for example, supposedly influenced a university, and other such accusations. Um, I'd like to ask you, Janette, is AI currently making everything worse, or is this an old problem that is just getting bigger right now? How do you see it? Well, for one thing, I agree with Eva that what one can say about AI has a very short half-life, right? Because the output is constantly improving, but what I saw from this supposedly Russian campaign was so poorly done. Then they put a blurry logo of a quality media outlet on top of these blurry videos and images. Then the text is in English , uh, which isn't necessarily the preferred language of their target audience. So what they want to achieve with that, besides saying, "Hello, we are currently running a disinformation campaign," is really not clear to me. I mean, you won't convince anyone with that. I'm relatively sure of that. It's always claimed that such campaigns are aimed more at destabilization, but I can't even see those effects in such campaigns. They are really not good. But nevertheless, one has to ask the question: in the long run, does it make a difference whether disinformation is produced by humans or by AI? And I dare to doubt that. We now also have this new labeling requirement, meaning that AI output must be marked as such. I believe that this regulation won't survive long, because in the medium term everyone will use AI at least as support, which is why it's more of a gradual question than one of "a human did this" or "AI did this." So, we definitely want to come back to the labeling requirement in a moment. I'm very curious. Do you see it that way too, if one really can't debunk it, or if one can identify it as potential disinformation because it's simply poorly made? Is it therefore automatically harmless; can anyone spot it? Yes, well, I think there are certainly cases that are so bad that we perhaps don't need to worry too much about the impact. But last year there was "Storm 15 16" or "15 16"—I'm never quite sure how to pronounce it— a comparable campaign, these " Doppelgänger" campaigns, and that one was a bit better made. And there, I do think we should be concerned that this trust in media, this fundamental skepticism, is being incredibly fueled. And when we talk about AI-generated content, such supposed news content, then I would like to draw attention to a completely different phenomenon. There are tens of thousands of AI-generated fake news sites that aren't there for us to look at and consume, but are used as a data basis for chatbots, for search engines, for these so-called hybrid overviews. That means the goal is no longer that I am personally unsettled by content that is a fake, specifically an AI-generated fake. That is not the goal at all; rather, the goal is for Perplexity, Google, and so on and so forth, OpenAI —meaning ChatGPT—to use this as a data basis to give me my personalized answers. And I believe at the very least—and we’re not talking about two or three websites here, we’re talking about thousands of AI-generated websites with just this purpose. Usually not hosted in Germany, usually not hosted in the EU, and all of that should—um—that should really alarm us. And well, at the very least: This blows my mind. Because suddenly we’ve reached a meta-level that is barely penetrable, except—and this is the same old song—through prevention, meaning we are fundamentally hesitant when we get an answer spat back at us, and we try to verify the sources ourselves; the good old song of digital literacy. So, disinformation with AI chatbots as the target group, which then, so to speak, spit out what they found in their responses to me. How can one protect oneself as a user of such AI chatbots? Not at all. We all want to use AI chatbots, too—and it's perfectly fine that we do. But we should really internalize this healthy skepticism, this healthy pause, and say: "It's great that ChatGPT is spitting this out right now, but maybe I'll prompt it three more times and ask: Now give me the source and show me the original quote." And that is what I mean by the good old song of digital literacy. You have to do that anyway, right? I mean, hallucinations have always existed, and now the phenomenon of "AI Slop" is increasing significantly, and it is certainly reinforced by the fact that AI is currently feeding on itself, right, and taking the very slop it produces as data input again. Um, that is definitely a problem. Still, slop reaches people too, doesn't it? So one has to state that, and perhaps, Leon, I’d be interested in your take: how do you view it when you see the same disinformation again and again, for example, the same fact or similar facts that are closely related about a specific politician or the like ? Do I have—or is something changing in me? Does this fact perhaps appear more familiar or credible? Do you think that, on a broad scale, one does reach people who might end up thinking a little differently over time because of it? I believe it depends very much on the sources and the form in which this information is presented again and again. If that is often, for example, poorly made AI slop, then I would argue it has the effect of me thinking, yeah, okay, it's nonsense, rather than me starting to believe this fact, that fact. Um, if it's well-made, of course, that's something else. And in this context, I would finally like to foster some discourse here. Um, I find this point about the infiltration of chatbots to be the much big—the much bigger problem, because misinformation can be spread much more subtly there, while I actually don't agree that we would be sitting here in a year saying we can no longer distinguish the videos . Because at the end of the day, if you look at the last one or two election campaigns, we already have a slowdown, a slowing down of progress regarding video generation, AI slop, and so on. We still have to invest a lot of technical time to run a very well-faked AI slop campaign, while everything else we see—and also these, um, cartoonish ones, I think this video you spoke about with the families, or what the AfD or AfD-affiliated accounts liked to do, um—that is still clearly recognizable as AI slop, and that hasn't changed. Therefore, I believe the danger isn't that great. And if we can all quickly learn to distinguish that, um, there is still hope. Um, which is a different question for chatbots. Okay, so we can all train ourselves a bit. You want to react—we are just mixing up two topics right now. One is the topic of deepfakes. One is, so to speak, the topic of fabrication, something that didn't happen, and whether we will still be able to recognize that in the future or not. So you are, so to speak, more optimistic there, and I am more pessimistic. The other, and that was the example of this AfD campaign ad, was that even though I recognize it's AI-generated, it still has an effect. So those are, I think, two different points. It is, so to speak, not about pretending here that it is real; that was irrelevant. I can recognize, indeed, I can recognize that it is AI-generated, but in this manner, in this presentation, it is still capable of being harmful to minors or impairing development. And I believe those are two different topics that we have been discussing together here under the heading of AI. Now I am wondering, with all these developments, is it really about concrete influence in this very moment, or are these actors we see playing a long-term game in any form? Jeanette, I’m interested in your take overall—how do you see this? Is this something where actors are perhaps thinking long-term and wanting to change opinions over years with many posts, for example, or is it really about mobilization for elections or something similar in the coming years? Well, I would say two things about that. For one, perhaps we shouldn’t overestimate the ability of individual actors to steer or control things in this context, because nobody really knows if they can sufficiently control the training datasets of such systems. But I wanted to come back to the long-term aspect and mention that what one can read about in this context today is that, in the course of media change, we have had such discussions— like the fear of manipulation and the question of distinguishing between media-generated and real—many times before, very intensely. For example, when photography was introduced, where the question kept coming up: "Is what I see in the photo actually what happened , or has someone manipulated it?" And that dragged on for ages, because every media creation is also always an intervention in the object being represented. And at some point, there were conventions in the press about what a photographer is allowed to do with a photo and what not. But that took at least 40 years to resolve. I don’t think we need that long now, but we will certainly have to agree on various types of conventions in the medium term regarding how to deal with these media, what makes sense and what doesn't. But we are going through a phase of uncertainty now that is typical for media change in general. And in the long run, I wouldn't be so pessimistic as to think we are all being manipulated and no longer know which way is up. Instead, that will eventually be refuted, and we will incorporate this new medium into our way of communicating—which will, of course, change as a result, but just as we did with previous media. So a short phase of uncertainty, I would Yes, or even a medium-term one; I mean, we had it with television, and we had it with radio, too. These are the constants of media development, right? The euphorics on one side and the pessimists and apocalyptics on the other. With the book press, I believe that holds true as well: the great promise of democratization on one side, and others who see the negative. Um, I claim to be a professional optimist. Perhaps I’m not, after all. I need to check my self-perception when I see this group here, because on one point I am perhaps, or to put it differently, I am obviously more unsettled than all of you. Um, and that is, so to speak, how we even obtain our information. And of course, there have always been lies, falsehoods, and the like, and yes, we in journalism have had to agree on conventions and new conventions with every new media technology. That is all correct. I just believe that the way in which we have tools at our disposal today, and specifically the immediacy—namely, through chatbots and similar things, constantly available everywhere—means that for good reasons, and perhaps fortunately, we no longer have gatekeepers who classify the world for us and establish their own conventions. Those who now have to establish conventions are the ones who provide and prepare the information for us, and I’m not so sure we’ll get there that quickly, because being a platform or offering a chatbot is not a profession, unlike being a journalist or a publisher, perhaps. And that point truly unsettles me a bit. It seems to me today that the space of information acquisition—the digital space—is permeated by different maxims than classic journalism may have been. Okay. And that brings up something I really want to ask about before we open up the floor, namely: what can we do, what is a possibility to perhaps get back in the right direction a little bit? Do we need new rules for social media and the providers, or where would you like to start? I believe we primarily need law enforcement. I don’t even know if we always need new rules and new laws and new regulations now. By and large, the legal framework is quite good. Be it with the Digital Services Act or the aforementioned regulation, including the effects we described at the beginning, be it with the European Media Freedom Act, or German media law. I could continue listing laws for another half hour, but I’ll refrain from doing so. So I don’t believe at all that new laws are needed, but rather effective application is what is required. The finest laws are worth absolutely nothing if they are not enforced. And in that respect, I am very glad to see that the EU Commission , which has taken on competencies for Very Large Online Platforms, is now really implementing them. The proceedings are underway, the first results—they are still preliminary results, but that’s how it is. They are now on the table, and the platforms must make improvements. There is the possibility of substantial fines, and that’s when it comes to a crunch, that these are used not only in competition law, in the DMA, in the Digital Markets Act, but also in the Digital Services Act at some point, and then a change in behavior will certainly occur. And that is where I am an optimist again. The European market is a giant market. No matter which platform, no matter if they come from China or from the USA or or it really doesn't matter. They cannot ignore this large market. The purchasing power here is too great for that. The commercial business here is too promising, and they certainly want to earn money. Therefore, there are real levers, there are starting points. The only question is how quickly and how effectively the current legal situations are actually implemented. So, there are levers and regulations. Jeanette, I would like to ask you right away, how far should regulation go, and are these rights that exist there really sufficient? Well, I would say, as far as chatbots are concerned, one would certainly wish that they also observe things like citizens ' fundamental rights, i.e., that freedom of speech is maintained and that no censorship is exercised. I am thinking of a study that we uh that I saw recently. Meta's Oversight Board uh asked various chatbots uh critical questions about various political leaders, also from non-democratic countries. And it turned out that all chatbots across the board refused to answer critical questions or did not react appropriately to them. And this type of restriction of freedom of speech or the failure to reference existing information, I find that a problem. Especially when one, as a user of these bots, is not able to verify it oneself. So, I would certainly wish for minimum standards in the long term, especially as these chatbots become more and more widespread. And there I would say, for example, that what the DSA has to do, namely transparency— what the DSA imposes on platforms, namely transparency regarding such editorial curatorial interventions. One would wish for that from providers of chatbots as well, don't you think? Absolutely. Um, it would also be, I expressed that wrong, if the impression was created that everything is fine. I believe the mechanisms are there, and now we have to look, a bit like in good old broadcasting law. There are very, very old principles: transparency, non-discrimination, open access. That already applied to cable networks. The young people in the room here might not remember that, but these are very old principles, and they work, and they are also future-proof, not just past-proof. I believe they are future-proof, but regarding the addressees, we really have to look; we have already partially imposed them, so to speak, on so-called intermediary platforms. And the next step, I completely agree with you, are these hybrids, this mixture of chatbots and search engines, and what actually is that? And the DSA isn't there yet. And I—this won't surprise anyone as a representative of German, uh, media regulation—also believe that we cannot leave this solely to the EU Commission, because there are certain areas that must be independent and also organized according to national rules. And that is where the respective national law comes into play. We already notice today, we have conducted a proceeding against Perplexity, or are still conducting it, or however you want to put it. Um, that we are reaching our limits with the laws as they are today, and that is exactly the chatbot, that is AI-generated news sites like Perplexity and so on and so forth. Not everything is fine, but the principles, the good old principles, are the right ones. We just have to adjust them again with a view to the addressees. Yes, do you see it the same way, just briefly, before we open the floor again in a moment and of course want to hear questions? Um, yes, absolutely. I, I think regulat—I also think that the regulations that are in place for what is already there work very well. But I also think that it takes much too long, and above all, it is adapted much too slowly to new developments, as you rightly said. Um, and I think what might be coming on the horizon now, what perhaps hasn't been in view much yet, is that we will also have a democratization of chatbots, in the sense that they are increasingly— especially due to open source, um, um, um, so to speak, the economic pressure from Chinese models, which are more and more, um, um, on par with US models, but are always released as open source because they aren't quite there yet. As a result, we are increasingly getting into a situation where we will soon see that, uh, uh, yes, models will be able to run locally on a wider scale and thus, if in doubt, could evade regulation even more. Um, and one should perhaps also be prepared for that. So, a danger that still threatens us there now. Um, or perhaps also a hope, how do you see it? Are you an optimist or a pessimist regarding this development? Fifty-fifty, everything is still open. Then I would like to look around the room for a moment. Does anyone here in the front perhaps have a specific question about that, or on a completely different topic? My colleague will come and give you a microphone so that people on the stream and everyone in the room can hear you. Yes, thank you very much. I have a question for Dr. Flecken. Um, you mentioned that the fact of whether content is AI-generated or not doesn't play a role for you in terms of the legal perspective, which I can understand based on the example given. But I wonder, what about the labeling requirement? Who actually monitors the labeling requirement? It cannot be the case that—well, we see as of today, um, I can set on Instagram that my content is AI-generated, I have to do that myself. The platform doesn't seem to check for itself today whether that actually corresponds to the truth or not. And scenarios are already imaginable where I, um, play with that, so in whatever direction, I hide it or I pretend everything is AI-generated; I can associate different intentions with that. Who would check that? The obligation is there, and, well, the, so to speak, the legal obligation. Okay. Who would check that? Exactly. Exactly. So, the AI Act indeed applies, and it applies to this point as well. However, a few adjustments are still missing, so to speak, regarding the implementation into national law, specifically with regard to the responsibilities. And for the media sector, and that is to be understood broadly, so not just the broadcasting sector, it looks like the media authorities will be responsible, but that is actually still a point currently in political discussion: who is exactly responsible for what? We also saw that with the already discussed regulation on political advertising. There it was also, well, parts of the DSA, while others are the responsibility of the state media authorities. Um, but that is currently sorting itself out. But in any case, someone has to be responsible, because just believing, well, it will work out, is perhaps not something we should count on. Are you satisfied with that or do you have a follow-up question? Yes. Does anyone else have a question here at the front, I believe. You were first. Hello, Aron is my name. I give workshops at schools, including today in Marzahn-Hellersdorf, on cultural education, media literacy, the topic of AI, deepfakes, fake news, and also populism in social media. My question concerns a bit of what you, Janet, mentioned, taking a closer look at messengers, because if these platforms no longer run political ads or if we simply see what an incredible amount of content is posted there on political topics that is not from political parties or is proportionately from few serious sources, doesn't the platform then to some extent leave the space to these other people, influencers, and also these AI-generated ones, which are then displayed a lot, but there is very little content from political parties? Um yes, and the platforms are kind of getting out of it because they say it's too complicated for us. Do you see these effects, that this content is taking over on these platforms? That is what I observe. And then another question for the research side. Are effects perhaps being reinforced where things are increasingly focused on specific individuals or party loyalty is lost because less party-political content is visible in this public sphere? Those are my questions. I mean, perhaps address the first question first. Leon maybe, or Jeanette, you were also addressed. I could provide figures. Um, so. For example, in the last federal election campaign, we were able to measure that only 1%of the videos in feeds actually come from official party accounts, and something like— conservatively estimated—5 to 6%are political with a specific party affiliation, but just not from official party accounts. Therefore, it’s definitely not videos from official party accounts that make up the vast majority of political videos— presumably over 10%—because in the end, the classification is very limited if you only look at party affiliation. Um, those were the figures. I'm not even sure if that's so terrible . I mean, because parties can still advertise. What is regulated and where they have withdrawn is paid advertising , right? And is that actually, um, are we missing something because of that? I'm not so sure. I don't know, how do you see it? Yes, above all we have to be careful; political advertising doesn't just mean messages or bought reach from parties—it can also affect an NGO. Um, it can also affect the EU Commission itself, for that matter. The EU—well, that's actually almost the cynical part about this whole "I'm withdrawing as Meta from the area of political advertising," because the EU Commission, for example, has a funding program for building renovations, or whatever, I have no idea. Of course, the EU Commission itself also bought reach on Instagram to make this funding program for the renovation of old buildings in Berlin better known. They can't do that anymore. They can now, on their accounts—I don't know how many followers the EU Commission account has —um, they can report on it organically, so to speak. It is also a political message, if you like, but just without the purchased reach. That means this topic, the platform's withdrawal from the area of political advertising, is not limited to just parties or ministries. Um, it can also affect Greenpeace, BUND, all sorts of things. And in that respect, the impact is much, much greater than just beyond classic party politics, and yes, whether that's good or bad. And, um, maybe it's actually still a bit too early to determine what the actual effect of this is. Are there actors who actually no longer have purchased reach , but simply use their reach as is? I don't really believe in altruism on social media. In that respect, I am curious to see what the state of research will tell us in a year or two about where this form of discussion and communication might have shifted. I don't know. We don't know the consequences yet. Um, we'll just have to wait, maybe we'll have another panel in two years and talk again about what has changed. Or talk to the platform providers about it, maybe they know. Yes, maybe we'll invite one here and then we can ask them again. Does anyone else have a question right here at the front, I think, or was someone else earlier? I think the lady here in front was earlier. What I felt was a bit missing in the debate is the topic of digital dependencies, because we talk about the platform providers as if they were neutral entities. Um, but in the current political situation, we are talking about very large tech corporations, essentially from the USA, that collaborate directly with the Trump administration and have a relationship of mutual dependency where they support and strengthen each other, thereby also supporting the autocratic tendencies that are currently increasing very strongly in the USA through these platforms. And I believe that is an aspect we also need to shed light on here. So these are not neutral platforms; the content distributed through them actually supports these autocratic tendencies and undermines democracy. And when we talk about regulation, we can also see very clearly that there were sanctions from the US side against Thierry Breton as a member of the then EU Commission, who worked very hard on the Digital Services Act. Um, and likewise the threat of tariffs against Germany, if Germany or the EU strongly supports this regulation. So, I am interested in your perspective on this: what influence does it have on the debate that we in Germany, in the EU, are massively digitally dependent on these technologies, and what do we need instead to free ourselves from this? Yes, perhaps we need new platforms from the EU? There are quite a few waiting in the wings, without promoting any specific ones. Perhaps Eva. Um, well, a very important step would be that those who have to enforce the law do not also have to fear receiving a call from Mr. Trump two hours later saying that car tariffs are going up. What do I mean by that? We really need independent institutions to enforce the law. Uh, I am quite positive, really positively impressed by the EU Commission. A year ago, I wouldn't have said that they had not only started these proceedings but have now also led to results, that it's starting to see penalties imposed, and so on and so forth. Because, of course, Ms. von der Leyen—well, whether it’s her or anyone else, it doesn't matter—gets a call immediately after, with a slap on the wrist here and there , asking what that’s actually supposed to mean. So the consequences are palpable, that’s the case, and I also believe that the tone has changed. And if you demand three times, "I'm in favor of a digital levy for large online platforms," I have no idea if I'll even be allowed to enter the US in the near future. Yes, there really is a cultural shift there and a disparity on both sides of the Atlantic. That was certainly different two years ago. I firmly believe in that. And I would like to illustrate this call for independence with an anecdote that has nothing to do with the US, but one that I experienced myself. As the media authority for Berlin-Brandenburg, we took action against Russia Today four years ago, and there was a situation involving the German branch, the German channel of Russia Today, which launched a station here in mid-December 2022. They just stupidly forgot to apply for a license. But you need one in Germany, so we opened a proceeding. Two weeks later, the then-Chancellor and the Foreign Minister were in Russia, and there were major press conferences; the Chancellor stood with the State President, and two hours later, the Foreign Ministers of Russia and Germany were on stage. And both German representatives said: "You may find it terrible that we are restricting freedom of speech or engaging in censorship in Germany because we are taking action against Russia Today." Of course, we had a reason: namely, a violation of media law. "But we aren't the ones doing it; those weird state media authorities are, and they are independent and at arm's length from the government." And I can tell you, I didn't receive a single call from the Foreign Office or the Chancellery. "You must ban them now, and how should we proceed from here?" Why am I telling this anecdote? Because it makes so tangible the immediate value of truly separating law enforcement from those who are lawmakers. And that is an anecdote that I hope doesn't happen to me too often. But hopefully, it shows that this isn't just something for show and that we should keep it in mind in a dogmatic, abstract, European way. And this naturally applies to the USA as well, that oversight authorities there, especially when it comes to media supervision and oversight of our public communication space, should truly be independent. But that is something which has a very European tradition and is already equally internalized or practiced outside of Europe. Not just recently, but for a long time. To that point directly. Yes, I see one point a little differently. We can certainly be glad that the EU, the Commission, has shown such commitment regarding the regulation of platforms and digital business models. But there is some weakness when it comes to enforcement and implementation. We see now with the digital omnibus that some things are being scaled back again. I cannot prove it, but many people see a connection there to the pressure coming from the USA. But this afternoon, there was an article in the New York Times that really impressed me. There has been an out-of-court settlement reached between Meta and many litigants. And Meta is now paying 13 billion, 17 billion euros for all the damages that have occurred in the area of youth usage. And that is only the beginning. My impression is that US courts might have more influence on the business models and conduct of the platforms than the DSA and DMA actually achieve. In fairness, those were all civil lawsuits. The other, well, is of course administrative law, so I believe those are two parallel tracks. Yes, of course. I just want to say, I was referring to the DSA and what these lawsuits in the USA are currently achieving. So, we need this pressure from various sides, not just in Europe but also in the USA, so that something changes. An exciting new point to conclude with: that this is now really being tackled from various fronts. Let's see who, in the end, actually brings a solution that will satisfy our group here. Unfortunately, our time for the official session here in the room is now over. There was a question here at the front, and one back there as well. I would really like to invite you to come up to us after the session and ask your questions during the unofficial part; we will still be here. You are welcome to come and talk to us. Otherwise, I would like to thank everyone who attended today. It was truly exciting, and of course, a heartfelt thank you once again to my panelists. Please give another round of applause for our panel. At our next salon on the last Wednesday of September, we will already know the election results because they will be over by then, and we'll know what the situation looks like. Nevertheless, I'll say to everyone in the room, um, we remain curious, and, um, have a wonderful evening. Have a safe trip home. The Digital Salon is a production of Alex Berlin and the Alexander von Humboldt Institute for Internet and Society. A new episode is released every month as a podcast or to watch on YouTube.