[00:00.000 --> 00:28.000] Good morning, everyone. [00:28.000 --> 00:33.000] My name is Hervé and I'm French. [00:33.000 --> 00:35.000] Sorry, my English is not very perfect. [00:35.000 --> 00:37.000] It's not my mother tongue. [00:37.000 --> 00:43.000] I work for an NGO called Open Factor, which is a French NGO. [00:43.000 --> 00:46.000] I'm going to talk about it right now. [00:46.000 --> 00:49.000] It's a self-funded NGO based in France. [00:49.000 --> 00:53.000] We created it in 2019. [00:53.000 --> 01:00.000] Our initial goal was to try to federate the Francophone Ozynt open source intelligence scene. [01:00.000 --> 01:08.000] As we noticed that the Anglophone one was pretty active and brilliant with Bellingcat, for example, people like that. [01:08.000 --> 01:18.000] So we decided to create this NGO and we wanted to assist the newsrooms and activists on Ozynt even investigation. [01:18.000 --> 01:28.000] We also wanted to train journalists about what we knew about and also wanted to promote young journalists based in France [01:28.000 --> 01:40.000] in order to help them to get some skills on Ozynt, which is a kind of a pragmatic way to find a job in France right now. [01:40.000 --> 01:48.000] As we are an NGO, a self-funded NGO, we also wanted to set up some philanthropic projects. [01:48.000 --> 01:56.000] We trained investigative journalists from Syria, which are relocated in Europe. [01:56.000 --> 02:00.000] We also did some programs in West Africa, the Francophone area. [02:00.000 --> 02:06.000] The NGO today is about 260 members and it's still counting. [02:06.000 --> 02:15.000] About me, some of the work first, we participated in some of pretty cool investigations, [02:15.000 --> 02:18.000] such as Greenblood for a media called Forbidden Stories. [02:18.000 --> 02:26.000] We also published some things and worked on things for BBC, for the French television on the Wagner group, for example. [02:26.000 --> 02:35.000] We also won a prize with Swiss TV for a documentary on war crimes in Syria, in Ukraine. [02:35.000 --> 02:40.000] Recently, we published two recommendable reports on Infos. [02:40.000 --> 02:42.000] I don't know if you've ever heard about that. [02:42.000 --> 02:48.000] It's a Russian company backed by the Russian intelligence. [02:48.000 --> 02:55.000] We also worked on that and tried to dismantle their network of disinformation inside their border. [02:55.000 --> 03:00.000] About me, I'm not a coder, obviously, as you probably see. [03:00.000 --> 03:07.000] I'm the co-founder of the NGO. I'm a former judicial investigator. [03:07.000 --> 03:10.000] I'm an open source software enthusiast. [03:10.000 --> 03:18.000] I used to compile Linux kernel since 1999 or something like that. [03:18.000 --> 03:22.000] I'm an open source software enthusiast and evangelist. [03:22.000 --> 03:25.000] I'm a techie, but not a coder. [03:25.000 --> 03:29.000] I consider myself rather like a Swiss army knife. [03:29.000 --> 03:35.000] I'm very lazy and believe me, if you want to solve a problem, hire a lazy guy or a lazy woman. [03:35.000 --> 03:37.000] I'm very curious. [03:37.000 --> 03:39.000] What about disinformation? [03:39.000 --> 03:41.000] This topic is about disinformation. [03:41.000 --> 03:51.000] Disinformation is, as you probably know, false information that is deliberately spread by an actor in order to deceive people. [03:51.000 --> 03:55.000] There is a fantastic researcher called Ben Nemo. [03:55.000 --> 04:03.000] He tried to resume this concept using the 4D's acronym. [04:03.000 --> 04:06.000] The first one is dismiss. [04:06.000 --> 04:08.000] The other one is distract. [04:08.000 --> 04:10.000] Distract the people, of course. [04:10.000 --> 04:13.000] Distort the truth and also dismay. [04:13.000 --> 04:17.000] So that's a very good way to define disinformation. [04:17.000 --> 04:28.000] It's an art for several actors. You can compare CTI and disinformation. [04:28.000 --> 04:32.000] There are some overlaps between these actors. [04:32.000 --> 04:34.000] There is a very great investigation. [04:34.000 --> 04:36.000] All the links are in the presentation. [04:36.000 --> 04:43.000] There is a really cool investigation called Doppelganger by the EU Disinformation Lab. [04:43.000 --> 04:52.000] I strongly recommend this report about media clones serving the Russian propaganda. [04:52.000 --> 04:56.000] I'm going to switch from time to time in the presentation. [04:56.000 --> 05:01.000] Disinformation is about war but also health, politics, economy. [05:01.000 --> 05:05.000] It covers mostly every topic, actually. [05:05.000 --> 05:15.000] Of course, internet and the web are probably the greatest echo chamber that they use. [05:15.000 --> 05:19.000] In France, we have such a problem. [05:19.000 --> 05:22.000] We have a problem with QAnon. [05:22.000 --> 05:25.000] There is a website called Cactus in France. [05:25.000 --> 05:27.000] It's called Cactus. [05:27.000 --> 05:35.000] It's for news and Cactus, the tree. [05:35.000 --> 05:40.000] QAnon is an American political conspiracy theory and political movement. [05:40.000 --> 05:43.000] I'm pretty sure you heard about it. [05:43.000 --> 05:46.000] It's mostly far-right based. [05:46.000 --> 05:50.000] It appeared on the net in 2017. [05:50.000 --> 05:55.000] This French website is, of course, Conspiracist. [05:55.000 --> 06:01.000] It's all about anti-vax and it's also very anti-semitic. [06:01.000 --> 06:03.000] It's a problem. [06:03.000 --> 06:11.000] It's present on several platforms such as, of course, the web, but also Telegram, Twitter, Gabber, etc. [06:11.000 --> 06:13.000] All the network. [06:13.000 --> 06:16.000] It has some kind of translated content. [06:16.000 --> 06:19.000] As soon as you arrive on a web page, it depends on the language, [06:19.000 --> 06:23.000] but you can get the news automatically in French or in English. [06:23.000 --> 06:26.000] This is via Tor, the Tor browser. [06:26.000 --> 06:31.000] It's basically in English, but you can also save it in French, for example. [06:31.000 --> 06:34.000] Most of the readers are French or Francophone. [06:34.000 --> 06:38.000] It publishes five or seven articles a day, mostly crap. [06:38.000 --> 06:42.000] It's always crap. [06:42.000 --> 06:48.000] It has more than two million visitors per month, so it's not nothing. [06:48.000 --> 06:50.000] It's a lot of people. [06:50.000 --> 06:55.000] I have a personal blog and I have an open-factor blog and we don't have two million visitors a month. [06:55.000 --> 06:57.000] I can assure you. [06:57.000 --> 07:02.000] Since it was open, it re-vendicates more than 80 million visitors. [07:02.000 --> 07:04.000] It's kind of big things. [07:04.000 --> 07:08.000] It's one of the most productive Q&A websites. [07:08.000 --> 07:12.000] In order to make the investigation, we needed a methodology on this information. [07:12.000 --> 07:17.000] There is a very good framework called the ABCDE, very easy to remember. [07:17.000 --> 07:23.000] It's a good start. A stands for actor, so who is doing the disinformation campaign? [07:23.000 --> 07:25.000] The behavior, how? [07:25.000 --> 07:28.000] The content, what about the content, of course? [07:28.000 --> 07:32.000] The degree, it means the scale of the disinformation campaign. [07:32.000 --> 07:34.000] And the last one, he. [07:34.000 --> 07:43.000] He was added later, but he means the effect that is looking by the actor. [07:43.000 --> 07:50.000] So if we based our research at open-factor for cactus, we would say something like, [07:50.000 --> 07:53.000] what is cactus? What's its audience? [07:53.000 --> 07:58.000] Its environment, ecosystem, its influence, its motivation, above all. [07:58.000 --> 08:03.000] And last but not least, if we can identify who is behind the cactus, [08:03.000 --> 08:06.000] it's really cool for us as journalists. [08:06.000 --> 08:13.000] So we decided to approach this by trying to qualify the environment. [08:13.000 --> 08:20.000] And in order to do that, we used three different tools that I mentioned as a sweet combo. [08:20.000 --> 08:25.000] The first one is HIFE, and it's a tool that was made. [08:25.000 --> 08:28.000] Most of these tools were made by the Media Lab Sciences Po, [08:28.000 --> 08:31.000] and HIFE is one of the greatest ones. [08:31.000 --> 08:38.000] It's a tool that you can download on your computer, and it mostly scrapes the web, [08:38.000 --> 08:44.000] starting from a URL, and you can build a web corpus of all the websites [08:44.000 --> 08:47.000] that are connected to this website by the links inside the article. [08:47.000 --> 08:52.000] So it was very cool to use with cactus, [08:52.000 --> 08:58.000] because the most important activity of cactus is writing articles, as I mentioned, [08:58.000 --> 09:02.000] but inside they put links to other articles, sometimes legit, [09:02.000 --> 09:06.000] sometimes links that lead you to other conspiracies websites. [09:06.000 --> 09:12.000] So kind of a long process, because right now, at the moment of the investigation, [09:12.000 --> 09:18.000] cactus had about 8,000 something articles on the website. [09:18.000 --> 09:21.000] So it took me a long time to scrape everything, [09:21.000 --> 09:27.000] and I had to go to the level two in order to scrape the link from the website [09:27.000 --> 09:32.000] that were cited by Kyaktu, if I'm clear about it. [09:32.000 --> 09:35.000] So I had to go at a level of two. [09:35.000 --> 09:38.000] It works pretty well with a caveat. [09:38.000 --> 09:44.000] There are some limitations with HIFE, especially one that really bothers us. [09:44.000 --> 09:47.000] It's Cloudflare, sorry for the typo, [09:47.000 --> 09:50.000] because as soon as you arrive on a Cloudflare website, [09:50.000 --> 09:54.000] and a lot of websites are protected by Cloudflare, you cannot scrape it. [09:54.000 --> 10:00.000] You cannot scrape them with HIFE, but it works pretty well. [10:00.000 --> 10:02.000] Actually, not on my computer. [10:02.000 --> 10:06.000] I tried at the hotel yesterday night, and it doesn't work, but believe me, it works. [10:06.000 --> 10:09.000] So I scrapped the links. [10:09.000 --> 10:11.000] I'm going to show you the result after that. [10:11.000 --> 10:15.000] The first tool that I combined with HIFE was, of course, Jeffy. [10:15.000 --> 10:20.000] I've been an enthusiast of Jeffy since probably 2011 or something like that, [10:20.000 --> 10:24.000] and it's a very interesting tool to explore and analyze graphs, [10:24.000 --> 10:30.000] and especially if you try to use the modularity algorithm to determine subcommunities, [10:30.000 --> 10:34.000] and it's a brilliant tool that I'm sure you already know about. [10:34.000 --> 10:37.000] And the last one was important for us as journalists. [10:37.000 --> 10:44.000] We tried to illustrate our articles with images and graphs, et cetera, [10:44.000 --> 10:46.000] and it's very difficult to render graphs on websites, [10:46.000 --> 10:51.000] so this tool is brand new, and one of its authors is in this room, actually, [10:51.000 --> 10:54.000] and I'm super happy to know that. [10:54.000 --> 11:05.000] And it's Retina, and Retina offers you the opportunity to import a graph inside your website [11:05.000 --> 11:07.000] and to dive into it and to analyze it, [11:07.000 --> 11:12.000] or at least to help the people try to understand it, and that's really cool. [11:12.000 --> 11:18.000] So that's what we got, and I'm going to show you the rendering of the article. [11:18.000 --> 11:22.000] So this is HIFE, and based on these three tools, [11:22.000 --> 11:26.000] what I tried to do is to illustrate the environment of the website. [11:26.000 --> 11:29.000] So the first one at the top level is Cactus, [11:29.000 --> 11:34.000] and all these websites were at least once mentioned in the articles, [11:34.000 --> 11:41.000] and then all the websites also cite the other one. [11:41.000 --> 11:47.000] And here, for example, you can reach the American disinformation network. [11:47.000 --> 11:53.000] You've got the Francophone one here and the Canadian one here. [11:53.000 --> 11:58.000] So it's very interesting to see where Cactus is in the middle of everything, [11:58.000 --> 12:01.000] so I tried to illustrate the graph like this, [12:01.000 --> 12:08.000] so it gives you with this three tool a very good idea of the environment of the website. [12:08.000 --> 12:13.000] What you can see inside also are some legitimate websites, [12:13.000 --> 12:16.000] such as Mediapart, for example, or YouTube, et cetera, et cetera, [12:16.000 --> 12:22.000] and that's one of their techniques in order to lead some activity on the website. [12:22.000 --> 12:28.000] They link regular websites and also conspiracies websites. [12:28.000 --> 12:34.000] It's in order to increase their SEO, their ranking on Google, for example, [12:34.000 --> 12:39.000] so it's very important to put legitimate and disinformation websites. [12:39.000 --> 12:46.000] So that was the first part of the investigation purely made with the open source software. [12:46.000 --> 12:51.000] The second one was the influence, and it's rather usual to do that, [12:51.000 --> 12:54.000] so we needed a good scraper for Twitter. [12:54.000 --> 13:01.000] Our idea was to have an idea on how the disinformation could escape [13:01.000 --> 13:06.000] from their traditional networks, such as Gabe or Parler, for example, [13:06.000 --> 13:10.000] and to escape from this platform to go to mainstream platforms. [13:10.000 --> 13:16.000] So Twitter used to be a very good platform to analyze. [13:16.000 --> 13:21.000] I don't know what's going on, and I don't know what it's going to be within a few days, [13:21.000 --> 13:23.000] but Essence Scrape is a cool tool to do that. [13:23.000 --> 13:25.000] I don't know if you heard about it. [13:25.000 --> 13:29.000] It's a scraper for Twitter. [13:29.000 --> 13:34.000] So what we did basically was to extract all the tweets from its creation [13:34.000 --> 13:39.000] that had a link referring to cactus.fr, which was the website. [13:39.000 --> 13:45.000] We wanted to know about its activity, so you just basically open a terminal [13:45.000 --> 13:47.000] and you copy-paste this. [13:47.000 --> 13:54.000] I'm not going to export to it first, but what you receive, I hope my connection is okay. [13:54.000 --> 14:00.000] Yes, so you get all the tweets, you can grab a cafe, you don't have to scroll, [14:00.000 --> 14:03.000] and when you come back, you get all the tweets. [14:03.000 --> 14:08.000] So there are approximately 124,000 tweets that talk about it. [14:08.000 --> 14:10.000] Once again, it's not nothing. [14:10.000 --> 14:16.000] And you can export that as a JSON file. [14:16.000 --> 14:17.000] Here. [14:17.000 --> 14:21.000] So it's going to work, blah, blah, blah, blah, blah, blah, blah, blah, blah, blah. [14:21.000 --> 14:24.000] I'm going to break it right now. [14:24.000 --> 14:29.000] Once you have these tweets, you have to scrape them and you have to wrangle your data. [14:29.000 --> 14:32.000] And I used another tool called OpenRefine. [14:32.000 --> 14:35.000] You probably know about OpenRefine. [14:35.000 --> 14:37.000] It's a powerful tool to cleanse. [14:37.000 --> 14:45.000] And there is a wow effect for journalists especially with this tool that I like. [14:45.000 --> 14:49.000] It's how it deals with JSON file. [14:49.000 --> 14:53.000] So this is going to be maybe possible. [14:53.000 --> 15:01.000] So you open the text like this and it knows how to render JSON file in CSV, [15:01.000 --> 15:06.000] which is a format that we all like when we do data journalism. [15:06.000 --> 15:10.000] So you can now analyze all the data that you have. [15:10.000 --> 15:13.000] So that's why I love OpenRefine. [15:13.000 --> 15:16.000] And also when you want to do some clustering. [15:16.000 --> 15:19.000] And the last one was, of course, R anti-diverse. [15:19.000 --> 15:21.000] I'm not going to dive into R anti-diverse. [15:21.000 --> 15:24.000] You have had several presentations and that. [15:24.000 --> 15:30.000] But these tools are really easy because R is an easy language for statistics, [15:30.000 --> 15:34.000] easy syntax, very readable when you don't know how to code. [15:34.000 --> 15:36.000] And this is really important for newsroom. [15:36.000 --> 15:39.000] For example, most of the time, future journalists know how to code. [15:39.000 --> 15:42.000] So having a code that is readable is important. [15:42.000 --> 15:44.000] You get a quick overview. [15:44.000 --> 15:46.000] So we've got 10 minutes, yes, thank you. [15:46.000 --> 15:51.000] It's very convenient for prototyping and important also, reproducibility. [15:51.000 --> 15:53.000] But if you want, you can use Python. [15:53.000 --> 15:57.000] I don't want to enter the war between R and Python. [15:57.000 --> 15:58.000] The motivation. [15:58.000 --> 16:01.000] The motivation, we use Firefox and we dive into the source code. [16:01.000 --> 16:09.000] And this one was pretty easy because QA2 has a Google ad program inside. [16:09.000 --> 16:12.000] So every click that you made on this web page gives you some money. [16:12.000 --> 16:16.000] So this is the main motivation apart from the ideology. [16:16.000 --> 16:22.000] And about the identity, all the tools that we got from GitHub, [16:22.000 --> 16:25.000] Holyhead, Jint and Sherlock. [16:25.000 --> 16:30.000] These are command line tools that are packaged right now [16:30.000 --> 16:32.000] by a French company called APOs. [16:32.000 --> 16:34.000] But you can use them on your terminal [16:34.000 --> 16:37.000] and it will help you to dive into the nicknames [16:37.000 --> 16:40.000] and the mail combination on websites. [16:40.000 --> 16:45.000] We also used some leaks, but we found out who was Patrick behind this website. [16:45.000 --> 16:48.000] So let's wrap this up. [16:48.000 --> 16:54.000] First of all, I want to talk about the myth of a single tool in Hozind. [16:54.000 --> 16:57.000] Most of the time when I say to people, [16:57.000 --> 17:00.000] well, I'm a journalist or I work on Hozind, [17:00.000 --> 17:02.000] they say, oh, which tool do you use? [17:02.000 --> 17:03.000] There are several tools. [17:03.000 --> 17:05.000] There is no single tool in Hozind. [17:05.000 --> 17:07.000] And you've got to mix everything. [17:07.000 --> 17:12.000] About this investigation, this guy, Patrick, is a retired IT technician [17:12.000 --> 17:16.000] in a lost town in the middle of France. [17:16.000 --> 17:19.000] He's a completist, but as we saw, [17:19.000 --> 17:23.000] he's in the middle of a bubble of a completist influence. [17:23.000 --> 17:29.000] He earns approximately $4,000 a month just by writing crap. [17:29.000 --> 17:34.000] So it's very important for us and for the audience to know him [17:34.000 --> 17:37.000] and to how to him on the website. [17:37.000 --> 17:39.000] Patrick, we know who you are. [17:39.000 --> 17:41.000] That's my thing. [17:41.000 --> 17:43.000] Why do we use open source software? [17:43.000 --> 17:45.000] Because they are powerful, adaptable. [17:45.000 --> 17:49.000] They respect your privacy and repress your reproductability. [17:49.000 --> 17:52.000] I don't ever know how to pronounce this word. [17:52.000 --> 17:56.000] Most of the time, these tools are free, which is really important. [17:56.000 --> 17:57.000] And I insist on that. [17:57.000 --> 17:59.000] Junos, we don't have money. [17:59.000 --> 18:04.000] And when you go to Africa and train NGOs, for example, or journalists, [18:04.000 --> 18:07.000] it's very complicated to buy software for them. [18:07.000 --> 18:09.000] They don't even have regular machines. [18:09.000 --> 18:12.000] So having open source software is really important. [18:12.000 --> 18:13.000] So thank you, you all. [18:13.000 --> 18:18.000] If you code for open source software, it's really important for civil society. [18:18.000 --> 18:20.000] Thank you. [18:20.000 --> 18:23.000] The last one, collaboration. [18:23.000 --> 18:25.000] I didn't talk about collaboration. [18:25.000 --> 18:30.000] When you're in the room today, it's important for us to mention [18:30.000 --> 18:33.000] that collaboration on how our investigation is the key. [18:33.000 --> 18:37.000] First of all, by software, but also in order to share information [18:37.000 --> 18:41.000] and to share indicators of confirmation, for example. [18:41.000 --> 18:43.000] So data capitalization is important. [18:43.000 --> 18:47.000] And there is this project called OpenCTI that I hope we will try to impose [18:47.000 --> 18:49.000] in Europe as a tool. [18:49.000 --> 18:51.000] It's an open source tool. [18:51.000 --> 18:55.000] And we'll try to impose it as a standard with the disarm matrix. [18:55.000 --> 18:56.000] It's called disarm. [18:56.000 --> 18:59.000] All the links are in the presentation in order to get some cooperation. [18:59.000 --> 19:01.000] Thank you very much. [19:01.000 --> 19:12.000] Thank you. [19:12.000 --> 19:20.000] Let's spare a thought for Patrick and on with the questions. [19:20.000 --> 19:21.000] Yes. [19:21.000 --> 19:23.000] You mentioned you used some leaks. [19:23.000 --> 19:27.000] Is that a leaks that you have friends that gave you information? [19:27.000 --> 19:32.000] Or is it that you found things that had been put inadvertently in the public space? [19:32.000 --> 19:33.000] Second one. [19:33.000 --> 19:36.000] What we tried to do when we do something. [19:36.000 --> 19:37.000] Sorry. [19:37.000 --> 19:39.000] I mentioned the fact that we use leaks. [19:39.000 --> 19:44.000] And the question was, is this information that some friends give me [19:44.000 --> 19:47.000] or information that is publicly available on the internet? [19:47.000 --> 19:49.000] That was a question. [19:49.000 --> 19:55.000] The answer is, as OpenFactor is an NGO, we don't have friends. [19:55.000 --> 19:59.000] So we use public leaks. [19:59.000 --> 20:02.000] We dive on telegram, the darknet, et cetera. [20:02.000 --> 20:08.000] And we use basically keywords, passwords, and emails in order to make connection [20:08.000 --> 20:12.000] between nicknames and try to identify people like this. [20:12.000 --> 20:14.000] That's what we do. [20:14.000 --> 20:15.000] Thank you. [20:15.000 --> 20:20.000] We also have a sort of, I'm a former investigative journalist, [20:20.000 --> 20:25.000] investigator, so what I try to do is sort of judicial methodology. [20:25.000 --> 20:29.000] So we try to open everything in our investigation, [20:29.000 --> 20:31.000] and we mention all the information that we get. [20:31.000 --> 20:33.000] If we get a link from somewhere on the internet, [20:33.000 --> 20:37.000] we will explain where we got the leak every time. [20:37.000 --> 20:39.000] Transparency is important. [20:39.000 --> 20:41.000] Another question here? [20:41.000 --> 20:42.000] Yeah. [20:42.000 --> 20:44.000] Are you Jewish? [20:44.000 --> 20:46.000] No, I'm not. [20:46.000 --> 20:47.000] OK. [20:50.000 --> 20:52.000] The question was, am I Jewish? [20:52.000 --> 20:53.000] No, I'm not. [20:53.000 --> 20:55.000] No, no, no, no, no. [20:55.000 --> 20:57.000] I'm not Jewish. [20:57.000 --> 20:59.000] Other question? [20:59.000 --> 21:00.000] Yes. [21:00.000 --> 21:04.000] I mean, you mentioned like one particular person that you identified, [21:04.000 --> 21:07.000] but I mean, I guess there are more people. [21:07.000 --> 21:10.000] What is the scope of like the success of those information? [21:10.000 --> 21:15.000] We try to, so the question is that we identified one person [21:15.000 --> 21:18.000] behind this website, and there are probably more people [21:18.000 --> 21:20.000] and more website like this. [21:20.000 --> 21:24.000] So we are just a bunch of cool people working behind their computer. [21:24.000 --> 21:28.000] So we try to do our best, but at least we have one website [21:28.000 --> 21:29.000] which is really influential. [21:29.000 --> 21:34.000] So we start by the most productive website that we want to investigate, [21:34.000 --> 21:37.000] but of course there are millions, not millions, [21:37.000 --> 21:41.000] but thousands and thousands of people doing these kind of things. [21:41.000 --> 21:45.000] We do what we can, and we try to inspire other journalists [21:45.000 --> 21:47.000] in order to make their own investigation [21:47.000 --> 21:51.000] and to out these kind of people. [21:51.000 --> 21:53.000] Sorry. [21:53.000 --> 21:59.000] How big is the environment of NGO working on this kind of subject? [21:59.000 --> 22:00.000] Are you the only one? [22:00.000 --> 22:04.000] Are you a lot of NGOs and friends working on this information? [22:04.000 --> 22:07.000] The question is, are we a lot of people, for example, [22:07.000 --> 22:09.000] in France working on this information? [22:09.000 --> 22:12.000] There are some newspaper and newsroom that start to investigate [22:12.000 --> 22:14.000] on these kind of topics. [22:14.000 --> 22:16.000] It's not very popular, not very famous, [22:16.000 --> 22:19.000] but I think that the last three or four years, [22:19.000 --> 22:23.000] this information has become a major issue, especially in Europe, [22:23.000 --> 22:27.000] and we see plenty of NGOs such as Hew disinfo lab, for example. [22:27.000 --> 22:31.000] They are friends of ours, and we really like them, [22:31.000 --> 22:33.000] and we try to collaborate with them. [22:33.000 --> 22:38.000] So, yeah, I hope that we'll get more and more people [22:38.000 --> 22:40.000] in order to work on that. [22:40.000 --> 22:41.000] Other question? [22:41.000 --> 22:45.000] How can you stop automatic disinformation? [22:45.000 --> 22:48.000] So I'm thinking about chat GPT and Vox and stuff like that. [22:48.000 --> 22:51.000] So can you distinguish whether it's being auto-generated [22:51.000 --> 22:53.000] or human-generated? [22:53.000 --> 22:54.000] So the question is that, [22:54.000 --> 22:58.000] is how can we stop automatic disinformation [22:58.000 --> 23:01.000] made, for example, by chat GPT? [23:01.000 --> 23:06.000] Chat GPT is going to be a very big issue for influence, [23:06.000 --> 23:08.000] which is something pretty obvious. [23:08.000 --> 23:11.000] When I'm saying that, I'm saying obvious thing. [23:11.000 --> 23:15.000] But, yeah, it's going to be a problem because [23:15.000 --> 23:17.000] websites, especially search engines, [23:17.000 --> 23:19.000] they use, as I mentioned, SEO technique [23:19.000 --> 23:22.000] in order to do the ranking of the website. [23:22.000 --> 23:25.000] And one of the biggest issues, for example, [23:25.000 --> 23:30.000] for Google, will be to figure out if a content [23:30.000 --> 23:33.000] not only is duplicated, which was pretty easy to do, [23:33.000 --> 23:37.000] but now automatically connected. [23:37.000 --> 23:39.000] So I don't have any solution for the moment. [23:39.000 --> 23:41.000] We haven't faced this problem so far. [23:41.000 --> 23:45.000] Chat GPT is quite recent, and its use is not [23:45.000 --> 23:48.000] widespread right now, but it's going to change [23:48.000 --> 23:50.000] within the next six months. [23:50.000 --> 23:52.000] And I'm sure about that. [23:52.000 --> 23:54.000] As an answer to this gentleman's question, [23:54.000 --> 23:58.000] the OpenAIG recently launched a site where you can [23:58.000 --> 24:01.000] give a text, it will give you a confidence level [24:01.000 --> 24:05.000] if it's human-generated or computer-generated. [24:05.000 --> 24:08.000] Yeah. [24:08.000 --> 24:12.000] Sorry, sorry. [24:12.000 --> 24:16.000] Yep, sorry. [24:16.000 --> 24:17.000] This one. [24:17.000 --> 24:19.000] Yeah, I was curious about the length that you did [24:19.000 --> 24:20.000] at the notes. [24:20.000 --> 24:22.000] Could you explain more of the color of the page? [24:22.000 --> 24:23.000] The level? [24:23.000 --> 24:24.000] Yes. [24:24.000 --> 24:27.000] So the level is basically the title of the website? [24:27.000 --> 24:28.000] Yeah, sorry, the coloring. [24:28.000 --> 24:29.000] Ah, the coloring. [24:29.000 --> 24:34.000] We use the, we use the, between the centrality [24:34.000 --> 24:36.000] on this graph. [24:36.000 --> 24:41.000] So it's usually what I do on this kind of website [24:41.000 --> 24:45.000] in order to figure out the importance of a node [24:45.000 --> 24:49.000] as a distributor of the connections [24:49.000 --> 24:51.000] between other sub-communities. [24:51.000 --> 24:53.000] So, yeah. [24:53.000 --> 24:57.000] Because the American cluster is, I'm just curious [24:57.000 --> 24:58.000] about your cluster. [24:58.000 --> 24:59.000] Ah, okay. [24:59.000 --> 25:01.000] Oh, the distribution, you mean, for example, [25:01.000 --> 25:02.000] of the links? [25:02.000 --> 25:08.000] It's, it's basically a, it's, it's the standard Gefi. [25:08.000 --> 25:11.000] No, no, no, no, the, the spatialization. [25:11.000 --> 25:12.000] What? [25:12.000 --> 25:13.000] For Satlas. [25:13.000 --> 25:14.000] For Satlas. [25:14.000 --> 25:15.000] For Satlas. [25:15.000 --> 25:19.000] So it's this one. [25:19.000 --> 25:20.000] Yeah, this one. [25:20.000 --> 25:23.000] So it's for Satlas, so every link that is close [25:23.000 --> 25:26.000] to the other one will be like packed as itself [25:26.000 --> 25:30.000] and a group, a cluster that is not very linked [25:30.000 --> 25:33.000] to another one will be far away from the other one. [25:33.000 --> 25:36.000] There is, there is a bias in the representation [25:36.000 --> 25:38.000] because I wanted QoT to be at the top. [25:38.000 --> 25:41.000] So it's part of my fault. [25:41.000 --> 25:43.000] But for the rest, it's basically the extract [25:43.000 --> 25:46.000] and the export from Gefi. [25:46.000 --> 25:51.000] Yeah, I was just wondering, in the beginning, [25:51.000 --> 25:54.000] you said that this information is deliberate. [25:54.000 --> 25:55.000] Yes. [25:55.000 --> 25:56.000] Incorrect information. [25:56.000 --> 25:57.000] Yes. [25:57.000 --> 25:58.000] How can you judge if it's deliberate? [25:58.000 --> 26:00.000] I mean, this guy behind the website, [26:00.000 --> 26:02.000] is he like, I'm going to push the slide [26:02.000 --> 26:04.000] or is he actually believing it himself? [26:04.000 --> 26:06.000] Is there a method to... [26:06.000 --> 26:09.000] First of all, when you work on this information, [26:09.000 --> 26:14.000] you, so the question is how can we make the distinction [26:14.000 --> 26:16.000] between misinformation and disinformation. [26:16.000 --> 26:18.000] So, basically, when you start an investigation, [26:18.000 --> 26:20.000] this is the first question that you ask yourself. [26:20.000 --> 26:23.000] Is it going to be disinformation or misinformation? [26:23.000 --> 26:28.000] Misinformation is like something that could be true, [26:28.000 --> 26:33.000] but no, that is false, but the person thinks it's true. [26:33.000 --> 26:34.000] Okay? [26:34.000 --> 26:36.000] But when you see the content of the website, [26:36.000 --> 26:39.000] you know that he is talking always on the same topics, [26:39.000 --> 26:41.000] always distributing the same kind of information, [26:41.000 --> 26:44.000] so you know that he does this deliberately. [26:44.000 --> 26:45.000] But the motivation is important, [26:45.000 --> 26:47.000] and that's why, in this case, [26:47.000 --> 26:50.000] it was important to identify the Google Ads tag, [26:50.000 --> 26:53.000] because you know for sure that its motivation, [26:53.000 --> 26:56.000] it's probably ideologic, probably, [26:56.000 --> 26:59.000] if you see his Facebook wall, for example, [26:59.000 --> 27:02.000] but there is the money, and the ads, [27:02.000 --> 27:04.000] and the Google and the platform [27:04.000 --> 27:08.000] are fueling the disinformation with the ad system. [27:08.000 --> 27:11.000] So, this is really important to identify. [27:11.000 --> 27:12.000] Okay, last question over there. [27:12.000 --> 27:13.000] Yes. [27:13.000 --> 27:16.000] Is there a strategy or a approach to tackling disinformation [27:16.000 --> 27:18.000] in authoritarian states? [27:18.000 --> 27:20.000] I didn't get the question, sorry. [27:20.000 --> 27:22.000] I mean, is there a strategy or an approach [27:22.000 --> 27:26.000] to tackling disinformation in authoritarian states [27:26.000 --> 27:28.000] and with partnerships? [27:28.000 --> 27:32.000] I mean, my information is far too beautiful. [27:32.000 --> 27:36.000] So, is there a strategy from, [27:36.000 --> 27:38.000] the question is that, is there a strategy [27:38.000 --> 27:42.000] to tackle disinformation from authoritarian states? [27:42.000 --> 27:44.000] I think so. [27:44.000 --> 27:47.000] The problem is that everyone has its own strategy. [27:47.000 --> 27:49.000] So, there are millions of strategies, [27:49.000 --> 27:51.000] and this is the big issue, according to me. [27:51.000 --> 27:54.000] So, Europe, for example, is trying to organize this [27:54.000 --> 27:57.000] by different programs. [27:57.000 --> 28:00.000] EU Disinfo Lab, like is the leader [28:00.000 --> 28:04.000] in trying to federate the disinfo structure [28:04.000 --> 28:08.000] and activists in order to make wider [28:08.000 --> 28:10.000] and bigger investigations like this. [28:10.000 --> 28:12.000] So, I think, yes, there is a strategy. [28:12.000 --> 28:14.000] The problem is that for the moment, [28:14.000 --> 28:16.000] there are multiple strategies. [28:16.000 --> 28:18.000] But it's going to change. [28:18.000 --> 28:19.000] Thank you. [28:19.000 --> 28:20.000] Thanks. [28:20.000 --> 28:21.000] Thank you. [28:21.000 --> 28:22.000] Thank you. [28:22.000 --> 28:23.000] Thank you. [28:23.000 --> 28:24.000] Thank you. [28:24.000 --> 28:25.000] Thank you. [28:25.000 --> 28:26.000] Thank you. [28:26.000 --> 28:27.000] Thank you. [28:27.000 --> 28:28.000] Thank you. [28:28.000 --> 28:29.000] Thank you. [28:29.000 --> 28:30.000] Thank you. [28:30.000 --> 28:31.000] Thank you. [28:31.000 --> 28:32.000] Thank you. [28:32.000 --> 28:34.000] Thank you.