Transcript
Speaker: I'm Jane Williams. This is the Red Eye Podcast. You can hear our live broadcast on Saturday mornings from 10 till noon at 100.5 FM in the Lower Mainland.
Speaker: Coming up in this episode, Laurentism speaks with investigative reporter Zach Veshera about Surrey Speak, a BC-based website with articles generated by ai More and more websites that position themselves as news sites are serving up what observers call AI slop.
Speaker: Last month, a Surrey-based site published fake news about Merrill candidate William Azeroth and an AI-generated picture of a real murder victim. Zach Vesra is a staff reporter for the Investigative Journalism Foundation, and he dug into the content and ownership of Surrey Speak, and he joins me now to talk about what he found out. Hello, Zach.
Speaker: Hey there. Thank you for having me. How does Surrey Speak position itself as an online publication? If you visit Surrey Speak's website or visit its Facebook page where it commands more than 10,000 followers, it will tell you that it is a community-based news site focused on sharing news about real people and real things happening in the city of Surrey, a massive municipality that famously doesn't have that much local news going on.
Speaker: It would give you the impression that it is completely run by a team of real human journalists. What did Suri Speak publish about One City, Vancouver's mayoral nominee, William Azeroth?
Speaker: SuriSpeak landed on our radar because they published a story saying that William Azaroff, who at the time had just been declared the mayoral candidate for one city, had resigned in disgrace, essentially, over some vaguely problematic tweets. The news article that they released is immediately kind of suspect because it didn't actually say what the tweets were about. It said sort of vaguely that they had something to do with indigenous identity and racial politics, but didn't dive into it and just said that Azeroth had resigned and that his party had promised to do better.
Speaker: There were a few problems with the article right away. ah One is that it said he that Azeroth had resigned from a position as executive director. He has never been the executive director of one city. It's a position he doesn't hold. He also didn't resign from anything. In fact, in real life, he had just been declared their mayoral nominee, sort of the opposite of a resignation, as you would nor had the party issued any kind of statement about him or or there being any evidence that he had tweeted anything that was anything close to what had been insinuated in that article.
Speaker: The entire thing was just very evidently and and blatantly false. It came on our radar originally because some of the folks at One City actually contacted me I've done some reporting previously about dark money financing and AstroTurf campaigns. And their original theory was that this was some kind of orchestrated political hit, that this was something that was designed to lower the reputation of William Azeroth, or maybe to put some negative news online so that in the future, if you Google his name, an AI-generated summary of that search might spit out some information about this.
Speaker: But as we dug into SuriSpeak, We reached kind of a scarier conclusion, which was that everything on this website appeared to be generated by artificial intelligence, not just the Azeroth story, but every piece of news there.
Speaker: A lot of these stories are riddled with factual errors, ranging from things that were just incorrect or details they got wrong to stories like the Azeroth story that are just complete hallucinations with no basis in reality.
Speaker: Can you give us a couple of other examples of false information published on the website? Totally. A more innocuous one was a story about a recent labor decision that came out about an Amazon warehouse here in BC that Unifor is trying to unionize. Those workers you know argued successfully that they were unfairly denied a pay bump that was given to other Amazon workers who are not unionizing, but the story speak ran a story on them basically saying that they had won a pay increase through collective bargaining. Now, if you're part of a union or you know anything about unions, you'll know why that's kind of a false insinuation. Collective bargaining suggests that these union workers have a contract with Amazon And they don't. that's That's one of the things they're really upset about is that Amazon has not signed a deal with them yet.
Speaker: And so it put this kind of false information out into the world. and And you could look at that and be like, well, maybe this is a writer who really doesn't know anything about labor relations or union law or how that works. Fair enough.
Speaker: But as you dug into the site, there are some way more problematic examples of of false information they put out. they They ran a story that was based on a real news item about an elderly woman in New Westminster who was recently robbed. She had her jewelry forcibly taken from her. And it ran an image of her that was AI generated. The image is very clearly based off of a real sort of video of the robbery in in and action. But it portrays a woman who does not at all resemble the real woman who was robbed. We checked this with her son to make sure. She's not even you know the correct ethnicity or height. like it's It's a completely different photo and it's very obviously an AI generated image as well.
Speaker: there was another case where we there was an image posted of woman named Melissa Blumke. She was 25 years old. She was murdered a few years ago in a very prolific and tragic way in Metrotown Mall. um There were recent news items about the sentencing of her boyfriend who was convicted for her murder.
Speaker: And the image posted by the Surryspeak of Melissa Blumke was also very clearly AI generated. it looks a lot like real images of Melissa, But again, we checked with a member of her family and we compared the images and ran it through a bunch of ai image detectors. And it's very clearly ai generated as well.
Speaker: Is there any transparency on the SuriSpeak site as to whether its content is AI generated or written by journalists?
Speaker: At the time we wrote this story, there was no transparency whatsoever about StorySpeak's use of artificial intelligence. All the articles that ran on StorySpeak had a you know a person's name attached them. Mind you, sometimes it was just only a first name or a first name and an initial, which if you've worked in journalism, that's an immediate red flag, right? Normally, the reason we have a byline is so we are identifiable and we can be contacted in in case there's a mistake. But nowhere on the website does it disclose ownership, who's actually involved, who owns it, or the fact that it was using artificial intelligence.
Speaker: After we ran this story, we noticed that SuriSpeak does begin to identify in some places that it's using AI, most notably on when it's generating images. What did you find out about the company's ownership?
Speaker: So SuriSpeak, as mentioned, doesn't identify who owns it. It's kind of a black box. But we were able to kind of follow some digital breadcrumb trails to figure out who was behind it. We looked at archived versions of the sites. We ran some detailed kind of forensics on the phone numbers and email addresses they used. We pulled corporate registration documents. And we were able to eventually trace, story speaks, ownership to a group called ADG Tech.
Speaker: ADG Tech is an Indian and Canadian tech company. It's incorporated in both countries and it specializes in applications of artificial intelligence. It's received backing from SFU Venture Labs, which is a tech accelerator affiliated with that university. And it it appears to have delved into the use of artificial intelligence to generate summaries of news reports. We were able to find another sort of new website that ADD Tech is openly affiliated with that is also sort of creating these AI-generated news summaries. And that website appears to be also affiliated with SuriSpeak. We also contacted a director of ADG Tech who confirmed to us that another director named Anuj Sayal was someone who took care of StorySpeak, was involved in its administration. I will note that after we brought this to ah ADG Tech, they denied having editor any editorial influence over the website. We then sent them a very detailed list of questions, kind of laying out all the evidence we had that connected story speak to them. And they did not respond to that list of questions at all. They have not contacted us or responded to any of our questions since we published the story.
Speaker: Well, where would you position that kind of lack of transparency about ownership, about who is running editorial, about who you can contact if there's an error for a site that characterizes itself as publishing news?
Speaker: It absolutely flouts pretty much every traditional norm of journalism, right? Typically, the sort of gold standard for journalism is that you know who runs a website, you know who's on the editorial masthead, you know who wrote a given piece of journalism, and there's a public and transparent process for contacting them in the case of an error.
Speaker: That is sort of the standard that we as journalists expect, but I think increasingly it's one that the public either doesn't expect or doesn't know how to look for. I think there's been a general corrosion of journalistic norms over the past like couple of decades in general. There's an inundation of false and misleading information online. And I think increasingly, a lot of people are getting their news from places like short-form video or just social media, where they're not necessarily doing the follow-up check of being like, okay, where is this information sourced from?
Speaker: Who wrote it? Who is behind it? And I think that's the kind of space where we see these AI-generated news sites cropping up. SuriSpeak is not an isolated phenomenon. We know for a fact that you know there's one organization that has cataloged, they say, more than 3,000 news sites that are you know posting AI-generated news, usually without disclosing that they are doing that.
Speaker: We have found other examples of this operating in Canada today. Very clearly, we are entering kind of a new phase of AI-generated slop content news. Zach, you quote Angela Misrian, associate professor of journalism at Toronto Metropolitan University, and she says, people don't understand that news is not the same as content. These are just content farms of garbage. What concerns is she raising?
Speaker: So the concerns that misery raised, I think, are actually really important for everyone who uses the Internet to be kind of appraised of. um I think that we take it for granted that we have like a consistent catalog of truth through published online material from reputable sources. And I think.
Speaker: There is a world where these AI sites kind of threaten that continuum. If we have a bunch of AI news sites that are pumping out flawed interpretations of the truth, you know articles or halluc articles that are riddled with errors or that are just outright hallucinations, there is a risk that eventually ai engines, like the one used by Google, will start relying more on those news sites generate their own AI-created summaries.
Speaker: And then those summaries will be fed into more AI engines that are relying on that. And you end up with this almost like this Ouroboros, that AI sites could just be feeding themselves increasingly flawed versions of events, like a photograph of a photograph of a photograph of a photograph.
Speaker: And we could get to a point, misery believes, where a majority of the things on the Internet are no longer true or don't have that basis in fact. And I think that is the scary that is like the scariest possible outcome of all this.
Speaker: Wow, that ah that sounds truly crazy. Now, you spoke with Demetrius Demetriatus, the director of research and development at the New York-based company NewsGuard. What did he have to say about why people run these AI-generated sites?
Speaker: I thought this was one of the most interesting parts of this entire investigation. As I mentioned before, one city's initial hunch about this was that this was a politically motivated hit job against William Azeroth, right? And they were sort of looking for the smoking gun of, well, who would be tempted to do this? But when I spoke with William Azeroth, he made a really good point, which was that that you know it was such a clumsy hit that it almost felt not sophisticated enough to come from a genuine opponent, right? Like it's a totally fake article. It's very easily disproven.
Speaker: And so that leaves open the question of like, well, why are people doing this if it's not to sort of torque political events and and influence folks? And what Demetrius's theory was, was that a lot of these AI generated news websites are set up purely for a profit motive. We don't think of journalism as being a particularly profitable industry. um I can certainly attest to that.
Speaker: But when you think about it, if your site is set up in such a way that it's just taking articles published by legitimate journalistic outlets and creating AI-generated summaries of them, essentially, and then spitting that out with an AI-generated image that's just poached from the internet, and it's just set to do this automatically using artificial intelligence, your overhead is essentially zero. You're you're spending very little money on the actual production of news because you're not producing any news.
Speaker: And that means that you can theoretically, if you get a little bit of ad revenue, if you're just getting enough clicks to your site, or you get a few advertisers to agree to advertise on your site who maybe don't have qualms about how you're producing your news, you're basically collecting straight profit. You're, you're, you're you know, you're, you've got a steady sort of stream of income that way.
Speaker: And that is what we suspect is the profit motive or the motive behind a lot of these news sites. There are definitely some AI generated sites that probably have more of a an ulterior motive. They're telling their AI to present a skewed version of events according to a certain political philosophy or alignment.
Speaker: But a lot of these are just ways of collecting easy money. Zach, what are your key takeaways for people using these you know so-called news sites and especially you know these AI-generated sites and perhaps more importantly, then going on to share the information that they find there?
Speaker: ah Gosh, I mean, I think it would be the same advice I would give to anyone anywhere about consumption of media and consumption of news. I really think that you should think about news the way, almost the way you kind of think about food in a way. You should think about like, you know where is this coming from? Is it good for me?
Speaker: Can my body digest this? Is it safe to eat, right? You wouldn't take something out of the fridge and it smells rancid and then put it in your body, right? And I think you should think the same way about the news and the information you find online. You should check out where it's coming from, who wrote it, what is the orientation of the outlet? Is this a reliable source of information? And just like with food, you know, you wouldn't offer your family, your friends, your partner, a piece of very obviously rotten meat, right? You wouldn't do that to them. Presumably you wouldn't poison them.
Speaker: Think about information the same way. Think about what you were sharing with people in your circle and try to put things forward that you know to be grounded. In fact, that is my general advice to anyone when it comes to sharing things online.
Speaker: Well, it's it's been really good to talk to you about a very alarming topic, but ah but I really appreciate the work you did investigating this. It's my pleasure. And if there's one takeaway from this, it's that local media is awesome. So thanks for what you do as well.
Speaker: Thanks so much, Zach. I've been speaking with Zach Vesera. Zach is a staff reporter for the Investigative Journalism Foundation. You've been listening to the Red Eye Podcast, produced by an independent media collective based in Vancouver.
Speaker: If you enjoyed this episode, it would be great if you left a review on Apple Podcasts. It helps other people to find us. Also, please consider sharing this episode with a friend or telling someone about our podcast.
Speaker: To keep up with the show, you can subscribe to Red Eye wherever you listen to your podcasts. And you can follow us on Blue Sky at Red Eye Radio.
Speaker: I'm Jane Williams. Thanks for listening.
