These news sites are really AI-generated Russian propaganda. Here’s how to spot them
Vatica Bell at the Big Board, good evening to you. With technology advancing as rapidly as it does, you're here to give us, and we all need it, a little lesson on how to spot fake news for ourselves. There's been a rising course of the last few months of Russian linked disinformation sites that look very misleadingly like real news outlets. Vatica tell us more. That's right, Mark. Well, let's take a look here we have on the left the Boston Times and we have on the right the Houston Post. And at first glance, these both might seem like legitimate or at least believable news websites. However, they they both are entirely resoundingly fake. These are actually disinformation sites. Those are, that is to say, sites that deliberately use deliberately with the intention to mislead. And they both use artificial intelligence to create their content and share a pro Russian narrative. In fact, this Boston Times outlet was even mentioned in this report here by the European External Action Agency as being part of something called Operation False Facade that is being part of a network of 23 inauthentic news websites that share pro Russian propaganda. And one, one example of this pro pro pro Russian stories we actually did cover here on France 24 a while back. It's this story here about foreign mercenaries in Ukraine committing atrocities. Now, if we go back to the Houston Post, there now the misinformation monitoring outlet News Guard confirmed to us that this website actually belongs to a network of 163 news websites, fake news websites that are managed by this man here, John Mark Dugan. And he's very much at the helm at the centre of this fake disinformation news outlet sharing empire, if you will. But as you can see, Mark, these are all the both of these websites are without a doubt not true. So Vedica, how can we all determine for ourselves what is what isn't trustworthy? Give us some tips on that. Yes, definitely. With a few tips and tricks, we can help determine whether a website is unreliable and whether AI is involved. So let's talk about the most obvious ones first. You may be able to find the AI prompt, the AI instruction somewhere in the website itself. Now, this won't always happen, but sometimes these do. These do slip between the cracks. So let's take a look at this headline here about the son of a Playboy model. In the headline, it says this title is short and precise, avoiding sensationalism while still conveying the necessary information. Now that is absolutely not intended to be there. This is definitely an AI prompt and that's that's what was fed into the AI to create and produce this article. Now, there may also be fakery visible in images. So if you could take a look at this, I don't know what you're reading there, Mark, but I definitely don't know how to pronounce that word here, truth and unpronounceable word. But this Boston Times logo was likely created by AI because of the inconsistencies. Now it's meant to, I assume it's meant to be some sort of impression of Latin, but these words don't mean anything. These are essentially gibberish. And you do see this a lot with AI. Sometimes it creates, it's able to generate the image that you've asked it to, but unable to to link those words together in context with that image. So if you do see things like this, there is a good sign that it is fake. Now lastly, you might be able to catch the AI hallucination. Now what is a hallucination, I hear you ask? Well, we spoke to Sheen Levy, who is the managing editor of News Guard, and she told us that AI work statistically and is trained on a database of word so that when they generate text or generate a sentence, the AI is determining what word is to statistically the logical next word in that sentence. But it's not able to process that data for itself. So when it creates false, nonsensical information and presents it as fact, that is what we call an AI hallucination of I'll show you an example here. This is an article in that website that we that we saw there, the Houston Post, where the AI has confused two different women from two different cases from two different time periods in the text. So we can see here that the AI has talked is talking about a woman who I found out is called Mary Berglund, and she's linked to a case from 2011. But in the same articles also speaking about someone called Madison Bergman, who is actually from a case in 2024, That's the new story that she was linked to. So two completely separate, separate people. But those are just a few of the things we should look to be spotting, Mark. It's an absolute minefield, isn't it? Now they're kind of like visual cues, I suppose. Are there more obvious methods for fact checking? What about finding more about the sort of fake source itself? Yes, that's definitely another straightforward way we can verify. Check the author of the article. Now, journalists on these fake sites tend to have fake profiles, fake BIOS attached to the website. Some of them go as far as to have fake social media profiles as well linked to the link to that bio. Now I'm going to show you one example here, the example of journalist Hector Payne, who has bylines in media alternative, which is a French version of these disinformation news outlets that I've been showing you. Now, Hector Payne allegedly also has bylines and writes for outlets like Le Monde and Media Paths and, and however, when I took a look into into Hector Payne, both of these are very well known, respected media institutions here in France. But when I take a look into Hector Payne, I couldn't find any trace of him online. There was no social media, no trace of him existing. There was no link to him at all on either of those news websites. And when we went a step further and actually contacted them to ask about him, they said that no one under that name has ever worked there. Now lastly, the about section on a news website could also be a pretty good giveaway. If we take this example here, The Houston Post has allegedly been in has has allegedly been going on since 1942. And some people may take that at face value. But if a paper with that name did exist, it was founded in 1932, according to the Library of Congress, and it ceased to operate in 1995. So if a paper did exist under that name, it definitely wasn't the one that we were looking at. But there you have it, Mark a few tips and tricks to help you determine which websites are unreliable and use AI.