You're listening to a stage talk titled Chinese influence operations and disinformation campaigns. This week we were joined by DoubleThink Labs Jasper Hewitt as we explored the new forms of propaganda. We discussed spam of large operations in the US elections to AI influencers pushing state lines on TikTok, along with some tool and reading recommendations along the way. This talk was hosted by me, Charlotte Ma, on Thursday the 8th of May 2025 in the Bellencat Discord server.
Well, for another stage talk. This week, we're speaking about digital influence operations and disinformation campaigns. An MIT study back in 2018, which feels like forever ago, found that fake news was 70 % more likely to be retweeted on Twitter than real news. And that was before Musk's takeover. So you can imagine what that
looks like now. While state -led influence operations have happened for years, but the use of internet, and in particular social media, as a means of running information operations has grown prominently over the past decade. The Chinese Communist Party is one such political power that uses digital influence operations and disinformation campaigns to advance the party's interests in the region and further afield. Jasper Hewitt joins us here
from DoubleThinkLab. DoubleThinkLab researches malign Chinese influence operations and disinformation campaigns and their impacts via the digital tools and methodologies they have developed. As an analyst, Jasper's focus has been on PRC influence operations during the Taiwanese and US elections, and he's currently investigating how official PRC accounts use the MetaAd library to expand their content's reach. We'll jump into all of
that in a second. But first, a short reminder, if you'd like to ask a question, please do so in the chat as we talk. And please note within your question, if you do not want me to read your username out, as this is being audio recorded and will appear on podcast platforms. Okay. Jasper, before questions, can you tell us a little bit more about your work? And while you do so, I will mute. Yes, so first of all, thank you for that great introduction and thanks for having
us, having Double Think Lab here. This is a great opportunity to talk to the whole community. So yeah, as she just said, Double Think Lab is a Taiwanese organization. We're an NGO and we've been founded in 2019 and yeah, we mostly conduct research into PRC influence operations. Now, I'm on the digital intelligence team, so I'm mainly responsible for tracking and reporting
on all of the inauthentic accounts. And we're also developing automated tools to track these accounts because, of course, this is a very time consuming work. So if we can automate it, then we do that wherever possible. But I can talk about that more in the Q &A. So yeah, today I just want to give you first say a few things about FEMI or influence operations and how that's different from fake news and disinformation. And then I just want to give you a crash course
into PRC influence operations. So we're going to talk about spamouflage, which is the biggest one. And I'm going to tell you about how spamouflage has been evolving over the last few years. Yeah, FEMI. So FEMI stands for Foreign Information Manipulation and Interference. It's a very long name, but it's basically a different term for foreign influence operation. Now the key term
there is foreign. So we're mostly concerned with information that goes from one country to another and then tries to influence the political debate in that second country. Femi is different from fake news and disinformation because not all influence operations are disinformation. So people can say something that is true, right? You can just say like, oh, I really like that political candidate. And that's not necessarily fake news, but we care mostly about the method. That's something
how information is spread. So we're looking for coordinated inauthentic behavior. So it's methods over over content. So if we find that there's like a thousand inauthentic accounts saying, for example, that BYD is an amazing company, then that would be a reason for us to to investigate. Now, if we look at PRC influence operations, there's several actors that are involved in this. So obviously, there's the PRC state media, right?
This is really big and they all have they're all very active on social media on several platforms. And then there's influencers. Now, some of these influencers have direct ties to the PRC government with some influencers. It's not as clear if they have ties to the PRC government, but people suspect that they do. And then there's spamouflage. So that's tens of thousands of, if not more, fake accounts that are basically just spamming the
internet with pro -PRC messages. So, yeah, spam has been defined as a massive cross -platform propaganda infrastructure. It was first discovered by Grafica in 2019. So Ben Nemo actually gave it the name spam because they would camouflage their propaganda with spam. So, yeah, you wouldn't only see propaganda. It would be like pictures of beautiful scenery or pictures of attractive individuals or random book quotes. And then suddenly
there would be this PRC propaganda. Now in 2023, Metta actually attributed spammaflage to PRC law enforcement. So this was a very important finding, but it basically it basically proved what everybody was already thinking, right? Because all of this stuff was so pro -PRC that there could really be only be one entity behind it. Now, what is the camouflage saying? Because this has also changed over time. At the beginning, they were mostly concerned with issues that are
sensitive to the PRC. So for example, prominent dissident Guo Wengui, he was targeted at the start, and then Hong Kong with the protests, COVID, and then also Xinjiang and Tibet. But over time, the campaign just became more overall pro PRC, anti -west. They started going after specific countries. So the USA is obviously a big target. Taiwan is a very big target. I think we're right at the forefront here in Taiwan. And then Japan, the Philippines, maritime disputes.
These are all topics that Spamouflage is touching on. So, for example, during the US elections, these accounts were also very active. And they were basically criticizing domestic issues, but also US foreign policy, like, for example, US support for Israel, for Ukraine. And they were critical of both presidential candidates. They're also going after individuals, so specific dissidents, also smaller dissidents, not just the super high profile ones. They're all being targeted by this
campaign. So their social media accounts, sometimes in the comment sections, you'll read a lot of hate comments. And those are also attributed to this camouflage. Now, what do these accounts look like? So I'm going to talk a little bit about how they operate. And then you can maybe also use some of that info to find these accounts yourself if you're interested in it. So many of these accounts, they would use profile photos
of women, mostly Asian women. And many of these photos have been taken from small Chinese influencers. So I personally think this is really messed up because these people, technically speaking, cannot use Western platforms themselves. But then an influence operation attributed to their government is taking their photos to put them out there on these Western platforms. The accounts mostly swarm around the larger state media accounts
and also influencers. So if you look at some state media accounts or accounts from diplomats, for example, Zhang Heqing is a big one. He used to be the diplomat to Pakistan, and now he's a big personality on X. And if you go through his posts and you look at his retweets, most of it will be, you will see his camouflage. Now, these accounts usually have a division of labor. So there's like cedars and then there's amplifiers.
So seeders would post the initial content, and then there's like 20 amplifiers below that, that their only job is just to repost that. We think that the accounts are side work accounts, so to say. So they're like hardly operated by scripts, but they're also definitely real people behind it. So we know that there's real people behind it because they are actually most active during PRC working hours. So if you... plot their activity, you will see that's from nine to five in PRC
working time. And then they also have a lunch break between 12 and two. So you can actually see that activity drop. So sometimes they'll also make robotic posts. So for example, sometimes at the beginning, there's robotic boot quotes when the accounts just start. And those also happen outside of working hours. Because obviously, when you run a script, there's no need to be
in the office. If you think of a fake account, then many people might think of an account that was just created the other day and has very little followers or something. But we actually found that many of the, specifically the Cedar accounts, they were created like back in 2012, 2011, and they have huge followings. So we found some accounts that have more than 70 ,000 followers, 50 ,000 followers. And it's not that The PRC was very
early with making all of these accounts. We found that actually many of them have been hacked. So those were inactive accounts that haven't been used for a very long time. And then they got hacked, and then they probably got sold to the people running Spamouflage. So one funny example with an account that we were tracking during the US election is when we finally get to get all of their posts through the API, we found that back in 2011, this account belonged
to a radio show in Texas. And then after that, it was just inactive for like 12 years. And then suddenly it became a spamflash account. We also found that these accounts appear to buy third party, like retweet amplification, especially on X. So during our investigation, we found over a thousand accounts that artificially boost these
spamflash posts. And these accounts were actually crypto accounts and Web3 accounts that normally speaking have a lot of, normally speaking, all of their reposts are of like large crypto accounts like ChainGBT or London Real TV. So it was really interesting to see that. in those timelines, suddenly you would see these spam of flash posts popping up. So there's some interesting dynamics going on there. It's not like they do everything
in house. There's definitely some collaboration going on with spam companies, potentially also from other countries. Now, spam of flash is notoriously ineffective. So many researchers say that spam of flash has failed to get a lot of authentic. authentic interaction over time. So yeah, what I just said, they buy their retweets, many of their followers and their retweets or other spam of flash accounts. So yeah, should we really
be caring about this? Well, today I'm here to say that, yes, I think we should keep caring about spam of flash because it is evolving and they are definitely trying to find new ways to become more relevant. So I'm now going to share. two or three examples of what we found in which we can see that spam of flash is becoming more
relevant, I think. So during the U .S. elections, they were continuing what they were always doing, like posting a lot of pro -PRC stuff and saying some critical things about the United States. But we also saw something new. And this was called MegaFlash. So there were these spam of flash accounts that were impersonating Trump supporters. And this was first discovered by Elise Tomas at the Institute for Strategic Dialogue. And yeah, there were these accounts that really pretended
to be American. So they had like in their bio, they had like all these American flags and said like stuff like hashtag NRA and stuff like that. They were interesting because it was harder to prove that they were connected to spamaflash. So whereas normally speaking, spamaflash is very straightforward. They just post a lot of pro -PRC stuff. So it's very easy to kind of see where their allegiance lies. But then with these accounts, they didn't post any pro -PRC stuff.
They just posted pro -Trump stuff, which was really weird. We were tracking a bunch of these accounts and we already had the idea that maybe these accounts were connected to the campaign because some of the crypto accounts were also amplifying these accounts. And then their posting times were also in PRC working times, even though they all claimed to be Americans, which is weird because that would be the middle of the night in the United States. And then, at some point,
they made this... So, Sprimal Flies is very clumsy. They sometimes make very clumsy mistakes. And then, at some point, like seven out of the 15 accounts that we were tracking, they suddenly posted this account in Chinese criticizing the Falun Gong, which was just completely out of context of what they were doing before. And I think after that, they kind of realized that the batch was compromised. They just left those accounts and they haven't been used since. So,
these accounts are still online. If you go through our US election report, then you can find the account name. Now, another thing, one specifically interesting account, it goes by the name of Edward D. Mayer. It's also still going today, even though we've reported on it many times. This account actually became really relevant. I think it's been the most successful spam camouflage account that we've been tracking over the last few months. So this account actually gamed the whole X algorithm
by posting Street Fight videos. So instead of boring spam, like some scenery, this account found that if he posts Street Fight videos, then the algorithm just boosts his posts. So he got some Street Fight videos for which he got like 9 million views. And this kind of spilled over to some of the propaganda posts that he would make. So on some of the propaganda posts, he
would get like 100 ,000 views. Now, this is a very weak spillover, but still a very big difference from what Spamouflage usually gets in terms of interaction. So with the other Spamouflage accounts, we maybe look at like 10 ,000 views, 15 ,000 views on a good day, and then 20 ,000 views would be like the jackpot, at least for what we check. But this account really managed to game the algorithm. and found a very interesting way to become more
relevant. At some point, this account got a warning label to its account from X saying that it was posting violent content and that influenced how it got recommended by the algorithm. After that, the account became irrelevant again. But he still continued like he created a few new accounts that I think were done by the same operator. And he also created like an X community, which was also interesting to see sort of trying out
all these new functions. And this account was also actively replying to authentic users with large following. So he was just like talking to some people and saying stuff about immigration and then people would actually respond to it and go into into dialogue with it. So, yeah, this was a very interesting finding, and it shows that Spamouflage, they don't sit still and stay irrelevant. They really look for ways to become
more relevant. Now, another thing that they've been doing is experimenting with artificial intelligence, of course. So I'm happy to talk more about that in the Q &A, because it's a big topic. But yeah, another thing that they've been doing is experimenting with new platforms. So there's a lot of new platforms popping up, right? Blue Sky, Threads, and especially Threads is huge in Taiwan. I think for now, for young people, Threads is probably the most important
platform in Taiwan. And Taiwan is going through some political challenges right now. Yeah, there have been a lot of influence operations from spam of flash targeted at Taiwan, specifically at the ruling party right now. So in this case, we also see a similar pattern. So also the spam of flash pattern, like spam and then propaganda. But it's more like the mega flash. It's not overly pro -PRC because they know that most Taiwanese people don't necessarily associate with that.
So it's more toned down. It's more like with the mega flash. They're only posting specific political messages that they think people might associate themselves with. And some of these posts actually already reached more than 80 ,000 views and 9 ,000 likes on the political posts. So this is very, very high compared to what we saw on X. This might also be related because the threats algorithm just works differently. And yeah, on Blue Sky, we've also seen a few
spam of large accounts popping up. And Blue Sky is particularly interesting because it has a really good open API. I don't know if anyone played around with it yet, but Blue Sky is just the ultimate platform to create an army of bots and just let it run. So yeah, but nobody is really using it yet or not enough people are using it yet. So if everybody switches to Blue Sky tomorrow,
that would be very interesting. And also, It's also a lot easier to do analysis on Blue Sky because the biggest challenge, I think, in this work has been that all of the platforms are closing down their access to researchers. So CrowdTangle disappeared, right? Got replaced by the content library, but it's not the same. And the X API is, yeah, basically completely closed off, right? You have to pay a lot of money. I think if you want to get a million tweets per month now, you
need to pay five thousand USD a month. And in the past, a researcher could get two million tweets per month for free. So, yeah. And TikTok is also a challenge in itself because the video format is just, yeah, kind of a pain in the ass to find information, especially because their search function is very incomplete. If you just typed like. if you do a keyword search, then it just gives you like 10 random videos or 20 random videos from random years that it thinks
you might be most interested in. So it's very difficult to do like exhaustive research there. So yeah, I think that's it for me and we can go to the Q &A. Amazing. Thank you. Yeah, I think most people within this audience or at least in the Bellicat staff side have experienced that frustration with the the tools closing down to monitor social media sites. Before going to audience questions, I was intrigued actually to ask you what the differences are, particularly as we've
got such a research -based audience. What's unique about researching Chinese disinformation campaigns in comparison, for example, with US state -led disinfo or Russian state -led disinfo campaigns? Are there any unique obstacles or anything that researchers who are used to looking at, for example, Russian propaganda might be surprised to encounter when it comes to Chinese propaganda. Yeah, I have to be really honest in that I don't know enough necessarily about the Russian disinfo
campaigns. This is something that I know that I should get more into, but I keep postponing it. I think perhaps the language element, the language aspect is very unique, maybe, to the Chinese influence operations. In the beginning, many of these accounts would post a lot in Chinese, but then now they're slowly using more English. Also, now there's AI, so it's more easy to just translate, oh, so there's less language mistakes.
But if you're looking at influence operations from the PRC towards Taiwan, which is interesting because they both technically use the same language, right? But then there's a difference in the characters that Taiwan and China use. So Taiwan uses the traditional characters and China uses simplified characters. look for in posts that we think might come from the PRC is we're looking for mistakes in the conversion in the conversion between those
characters. So many PRC threat actors, they will write a post in simplified and then they'll throw it through software to convert it to traditional characters. And then there's always mistakes there. So, yeah, we're kind of like always looking for these simplified characters that might give
us a hint. Yeah, absolutely. You can often find, for example, with Russian disinfo campaigns as well, language, even in translated English, can help you spot if something is potentially misleading or not, because often the language translation isn't incredible and, as you mentioned, is often based on translation bots online. We've got quite a few questions coming through. We've had a few, because you mentioned about the nature of social
media as well. Sia Skippy asked, has X's moderation of disinformation varied with the nation of origin and content of the disinformation? As in, do different countries get different and different governments get different treatments? And kind of mentioned with that, Ozkom said, you've mentioned that you've tried to get these accounts banned with little effect. What other mitigation opportunities do you see? So when you're looking at social media, Does moderation change depending on where
these disinformation campaigns come from? And how what other mitigation opportunities do you see when banning these individual accounts doesn't work? Okay, so I'm not sure if there's if there's a difference in terms of in terms of geography, like if you do it in this country, or if you do it in other country, what I did notice is there's a difference in time. So if we find an account now and we write on it, I really highly
doubt whether X will ever delete it. But when we were doing this research during the US elections, like in the weeks leading up to the actual election, and we had a collaboration with Voice of America. So Voice of America was just churning out articles about this all the time. And back then, we did notice that a lot of accounts that they reported on actually got deleted, like in the days after
the article broke. And in some cases where we really found that someone was being impersonated, for example, and then an ex -spokesperson actually came out to say something about this and said that they deleted that specific account. So I guess at that point, even ex itself realized that they had to do something. because otherwise there might be too much pressure on them that
they're not doing anything. And there's other accounts like, for example, the pay for repost accounts that I mentioned, like the crypto accounts, they were banned very, very rapidly. But I think that's just like an anti -spam campaign that they always have running and it's not necessarily related to disinformation. But like those thousands of accounts that we found back then, I think
all of them have been deleted today. So in terms of other methods for mitigation, if the platform doesn't want to take it down, I don't really see another thing that you can do. Maybe if a lot of people report it, maybe at some point they will do it. I also am very skeptical about that. So I think the best thing you can do is to just tell other people about it so that everybody knows that this kind of this badge exists and that people are aware that these kind of accounts
belong to an influence operation. And that's especially relevant right now for Taiwan, where so many people are on threats. And right now there's so many disinformation going on on threats and normal people are seeing it in their daily lives and are talking about it at work. Yeah, I think just calling it out and calling out that this exists can sometimes already be enough.
Absolutely. We have seen a pattern personally, though, at Billingcat, for example, and I know a couple of other news organizations have as well. When it comes to news impersonation, when there's tactics where they're impersonating news organizations, for example, calling it out has actually given these individual accounts more viewership and more engagement. So then they've
been using that as a tactic. And we've seen an outtake in, for example, in terms of the Russian disinformation context, people, Russian disinfo drivers, emailing and trying to get the attention of European fact -checkers, for example, so that they can fact -check the content to then drive more engagement their way. Interesting model that has been working for them and is something that we've spoken about in a stage talk before. We've got so many questions rolling in. I'm hoping
I can keep up with them all. Just keeping on the social media side and as you were talking about the spammer flage, particularly within the election that you spoke about, Skirskippy asked earlier as well, how many of those followers of these accounts are themselves legitimate or bots? For example, my experience on Twitter is that accounts just naturally accrue totally authentic
followers at a steady rate. Is there kind of a secondary group of bots that are being created to amplify those particular tweets or those particular posts? Is that part of the plan? Or are these real people kind of getting suckered in to kind of the propaganda and the messaging? Yes. So this is a very important question. And I don't know if you're familiar with the breakout skill.
So this was also developed by Ben Nemo. So the breakout skill, basically, when you look at disinformation or influence operations, you want to see like, did this post break out of its initial bubble. And the initial bubble is just an inauthentic bubble with just a bunch of fake accounts that are following each other and are liking each other. And then if it breaks out of that bubble, you would be in category two. And if it breaks out on multiple platforms, you'd be in category
three. And then it just keeps going up. So we're always trying to look for if these accounts are getting authentic interactions from real accounts. And it's difficult to look at all the followers. It's possible, but if you have these large accounts that I mentioned earlier, like these hijacked accounts, those are tens of thousands of followers. I think it might be somewhat of a waste of time to go through all of their followers. So what we usually do is we look at the comment section.
Because the comment section of spammer flash accounts are very, let's say, Most of the other spam and flash accounts that comment on these posts, they all agree. They all say stuff like, yes, yes, good point, stuff like that. It's very kind of like monotone. But if at some point you see replies that disagree with the post, or you maybe see replies with people saying stuff like, oh, this is weird, or whatever for some reason, then you can assume that this post has somehow
reached authentic users. So I think that's the easiest way to see that. Yeah. And then, yeah, just a little more on that. So I think they themselves, Spammouflage itself is creating more and more Spammouflage accounts to then follow all of these accounts and also repost them. And then on the other hand, I think they're also buying it from third party. So like spam companies, they're
also buying likes and reposts. You mentioned that in terms of the propaganda aimed at Taiwanese population, that's happening a lot on threads. We've actually had a question from Sohan D'Souza who's asked, would you please advise regarding optimal techniques for threads investigations? The advanced search doesn't seem too bad, but transparency even via external services seems
lacking. Could you speak a little bit about how the difficulties and the ease of investigating, particularly on threads, is that's a new platform that many of the researchers, for example, in the audience may be new to and want advice on. Yeah, sure. So I think one thing that's difficult about threads is that everything is new. So sometimes when people look for fake accounts on X because in most cases, it still are like new accounts,
right? So you would suddenly see an account that was created like three months ago saying all this stuff. But then like on threads, it's not weird to have an account that was created three months ago because everybody is just joining this year. But yeah, I think threads is available in the content library now if you have access to that, but it has a lot of limitations, like only certain accounts you will see. So it's not
very helpful. What I think is a very useful tool is there's this add -on that you can get in the Chrome store. It's called Single File. And it basically just lets you get a .html file of the whole page that is currently loaded. So you can just go to someone's timeline, and you can scroll all the way down, and you can click the add -on. And then it will just download a .html file of that whole. of that whole page. So that would
include all the posts, all the images. And what you can also do after that is you can write a script to organize that data into a data frame. So technically speaking, you're not scraping. You're not using Selenium and stuff. So Facebook won't detect you and ban your account, because you're just using this tool to get all of the HTML. And then you can do with that data what you want after. Another thing that's difficult on threads is I found that sometimes certain
things aren't really working. Like when I try to click the account creation date, for example, on my laptop, sometimes I get a lot of errors, but then when I do it on my phone, it works. So I feel like the app works better on your phone than it works on your computer, but I'm not doing investigations on my phone. And then you also have to do a lot of clicks to get there. So it just takes a lot of time. Other than that, I think I like that it has a keyword search function
with a recent tab. So I think it looks pretty exhaustive like what you have on Twitter. So I'm really happy about that. That it's not like TikTok. Thanks for that little tip. I've linked that in the chat, and I've personally just gone and downloaded it. I have a similar one called go4page, but that just takes a PNG, so it's good to know that there's one for HTML. To say more on that, this works for almost every website.
The only downside is that the .html files are very big, so if you get someone's full timeline, you might get 100 MB. But yeah, I think it's kind of like a half manual, half automated way for data collection. It's also a great way to archive, right? We all know the trouble of trying to archive something with your VPN on and then you have to, it just takes ages. But with this thing, you can just click it and save it for later. And then if the account gets deleted,
you can just go back to the .html and make your screenshots from there. And that's a great time for me just to shout that we have a Bellingcat auto archiver, which has recently had some updates done to it. So please do check that out as well. Just linked the guides that in the chat. We've had quite a few questions and I'll let you take
a... Choc? Cool name. It's the most recent question to kind of encompass this, but we've had quite a few questions about how you spot Chinese Finney or Spammelfarge accounts when their activity seems to be non -discernible from other Finney without an in -depth analysis. Are there any telltale signs to spot Spammelfarge accounts,
specifically Chinese Spammelfarge accounts? Yes, so I think some of the things that I just mentioned, like for example, that they're mostly active during PRC working hours, that is one that you can always check. And you can also look at the content. So for example, I mentioned a few things like specific dissidents or Hong Kong or Xinjiang. So these are like... target topics that they always focus on. So if it's like a traditional spam -a -flash account, it will talk about these
topics and it's very easy to spot. So you can just go to, for example, you can go to People's Daily's X account and then just go through its repost or look in the comment section and you'll find a bunch of spam -a -flash accounts. Now it becomes more tricky if the spam -a -flash accounts are like these mega -flash. type of accounts, right? So they're not posting all of this traditional pro PRC stuff. And you might think like, oh, this could be, this could be
anyone running this influence operation. So how we attributed those accounts to spamouflage is by looking at the people who amplify it. So that's where these kind of weird crypto accounts came in. So we noticed that these accounts were amplifying a lot of spamouflage and then suddenly like right before the election, like five days before the election, just all these new accounts were created that criticize specific lawmakers. And they only
had three posts and nothing else. But when we looked at the reposts, we saw that there was actually quite some overlap with the accounts that amplify the traditional spam of flash and also PRC state media accounts. So that's how we were able to attribute it. There's also more on that in our US election observation report, if you're interested. And I highly recommend anyone to just read a lot of papers about spamouflage, and you will see a lot of account characteristics
that you can use to find them. And once you find a few, you can just keep snowballing. Someone asked earlier actually in the chat, in the election coverage that you saw, were they... only promoting Trump rhetoric, or was it a little bit more broader across the political spectrum? They also asked, have you seen any similar campaigns targeting European politicians, for example? For example, last year was a huge election year across the world. Have you seen any similar trends in terms
of European elections? So first, on the first question, most of the traditional spam camouflage accounts were anti -everything. They did say a few things that maybe aligned more with Trump. So they were, for example, a bit more, well, they're anti -U .S. support for Ukraine, which maybe you think that, okay, maybe that's more closely connected to what Trump is saying. There were also critical of the LGTBQ movement. So maybe you can also draw some connections there.
But with the mega flush account, specifically the accounts that are really pushing Trump, we only found that for Trump. We didn't find similar accounts that were supporting Biden. One interesting thing, though, is that basically all of the spam flush accounts that we were tracking were very anti LGTBQ. And then like those five days before the election, when these accounts were created that criticize specific lawmakers, they were criticizing Republican lawmakers for being anti
-LGBTQ. So that's just a completely different standpoint kind of playing both sides. So that was it was interesting to see that. And for Europe, I haven't really been tracking the European election season and also the accounts that I was tracking, it was more quiet about the European elections, but I'm pretty sure that some stuff did go on. Maybe that's something our community can have a look into. Something to track, guys. A little homework for you all. You mentioned earlier that
often they use images of influencers. as part of the accounts, they might not be aware, these influencers might not be aware that their images are being used or it's being used without their consent or their control. Do you also see a trend of proxy actors or self -interested actors helping to boost propaganda narratives in Asia, particularly? For example, social media influencers peddling Chinese state lands. I think on your website, it might attend economic capital owners or suitors.
Is there any pattern to there to what these influences look like, say, or do they kind of transcend algorithmic communities if they do exist? Yeah, so one of our partners actually wrote a big article on that. That's also on our medium. And she found that on WeChat, so now you have these WeChat stories, and she found that there were a lot of influencers. Chinese and also Taiwanese influencers, like living in Taiwan, who would then post a
lot of very pro PRC narratives. So to say like, oh, Taiwan should join China and that kind of stuff. But the conclusion there was that she believes that they were only saying these things because they know that they might be boosted by the algorithm for saying that. And then they were all running these online shops on the side. And this is interesting because we found a similar
thing on TikTok. I think Voice of America also wrote an article about that, where we found these accounts that were kind of operated, seen from out of Singapore, and they were AI avatars, often of Western -looking women or Russian or Ukrainian women. I think that's how it kind of started.
And they would start by giving this pro PRC propaganda talk and then like halfway through the video they would suddenly start promoting these vitamin products and it's just so weird but there's this this theory that these videos were maybe originally made for PRC social media and they think like oh if we post this kind of stuff the narrative that the algorithm will push us because that's what That's what Xi Jinping wants us to say. And then we use that to then sell our products.
But this is still kind of like a hypothesis. But we did find this in multiple different places. That's really interesting. Yeah, I haven't seen it linked with political comment, but I've definitely seen news influences. fake ones, using a news hub to then go and sell a well -being product, for example. It's really interesting that that's a tactic that they might be using, particularly on TikTok. Thanks for sharing. And I've linked, I think, is it called Patriotism for Sale, the
WeChat one? Yes, yeah, that's the one. I've linked that within the chat. Do check out DoubleThink Labs' medium, by the way. It's full of incredible information. Hania Reyes just put, I've seen hacked influencer accounts start to sell products on TikTok too. Mashtar said maybe they modeled some of it off Alex Jones info walls as they were pushing vitamins. Somebody's just dropped your DoubleThinkLab medium in the chat as well. That's really cool. Thank you for doing that.
Still so many questions coming in. You mentioned the big... the big term AI in your last answer, could you tell us a little bit more about how AI is being used in influence operations going forwards and is that a new phenomenon or has it been going on for a few years? I think right now the thing that AI, that people are most afraid of with AI are not are not really happening yet, at least with PRC influence operations. It feels like they are kind of experimenting with it.
So I kind of divide it up for myself. I think AI is being used in four different ways. Now, the first one is AI that is being used to deceive, right? So you're trying to create photos or videos that look real, but actually they're fake. And you're really trying to fool people. I changed their perception of things through these fake videos or photos. And so far from the PRC influence operations, I have not really seen that. Most of the use of AI has been AI to convey a certain
idea. So they would, for example, create an AI generated image of, I don't know, like Trump and Netanyahu kissing or something like that. And that would be like everybody knows that that this is fake, right? It's very obvious, but it's being used to get this idea out there, get this visualization. I guess it's just a replacement for a cartoon if you don't have time to draw a cartoon yourself. And then there is another method, which is also interesting, and that's
AI as a technical assistant. So for example, we found Also, in the US election, we found that some accounts, like a few days before the election, they suddenly uploaded a video and it was like speech in English and then some stock images in the back. But we actually found that the speech originates from a Baidu post in Chinese, like an essay that someone posted. And yeah, we highly suspect that they used AI to translate and then also use text to speech AI to create the voice
to say this. So this opens up a lot of possibilities. Right. So they can now just get any any data from the Chinese cyberspace and just turn it into a spoken like perfectly spoken English audio fragment and just put it on Western social media. There was something similar in Taiwan during the election, so someone made a PRC threat actor wrote this book about Tsai Ing -wen, the previous president, and they called it The Secret History of Tsai Ing -wen, which is just a book full of
nonsense about Tsai Ing -wen. And then they got all of these AI avatars that they got made with CapCut, so I think that belongs to ByteDance, and they made all these AI avatars. to narrate the book in kind of like a newsroom setting. So yeah, just kind of using these AI tools to very quickly get certain information out there. And I think the most scary one, which is like AI for full automation, which is the thing that I am most afraid of, we've seen accounts experimenting
with it. Like we've seen posts, even posts by, for example, Zhang Heqing, like the former the former PRC diplomat. He has a lot of posts that look a lot like they were AI generated. And I think that that's where it becomes really scary because that's the moment where you can just create 10 ,000 bots and give them a wide range of prompts. You can give them guidelines and they just can create original content all the
time and it just won't stop. And especially because OpenAI, like in February, they released their newest report in which they found that PRC threat actors were using JetGBT to create critical comments of certain dissidents. Then you can just unleash an army of 10 ,000 bots on one person, especially if you have a model that you can locally run. This is what I'm most afraid of, but it hasn't really happened yet. Thanks so much for explaining some of the tactics that they take in terms of
AI. In regards to other questions, Mini Joyce has asked, what are the odds China uses cyber capabilities to bolster their Femi operations? I mean, more than they already do. Is there a sense where they're using cyber capabilities to bolster the reach of these kinds of operations? I think that is definitely possible at some point, especially if If you start looking at full automation that I was just talking about, I think at that point, they could definitely tap into some of
the cyber resources that they have. But I also think that you would maybe see this on creating news websites and stuff like that. I think you would see it would be more prominent there than on social media accounts. And yeah, of course, they have Yeah, you have the rise of PRC platforms right now. So yeah, TikTok and also Xiaolongsu, like Red Note. People have also started downloading
that. And of course, there's also a lot of, I think, embedded influence operation going on within these apps with pushing certain content and suppressing other content in the same way that happens on Weibo or like the really proper PRC platforms. So they're definitely using their cyber capabilities there. As you asked earlier, and this goes back to kind of people in the audience, for example, who might be interested in getting
involved in all of this. What research, what resources slash learning platform slash certifications would you recommend someone trying to get into the field of FEME? I think the field of FEME is kind of Well, it's maybe similar to OSINT in general, but the field of FMI, most of the analysts, especially at a double think lab, all come from kind of like a political science background. I also have like a background in Chinese studies
and Asia Pacific studies. And then I learned some coding and some data science, but we also have an engineering team that can really help us with more of the technical things. But I think just some basic Python, especially pandas, could be very useful. Because, for example, when we write algorithms to detect coordinated inauthentic behavior, we look a lot at, for example, post similarity. And the time between posts is also
very important. So if you can very quickly. kind of like manipulate the data to find similar posts using certain Python packages and also find posts, batch posts that were done in very rapid succession, then that is definitely very helpful. I guess if you're very good with Excel, you can also do all of that stuff. I don't necessarily think you need any certification. or specific certification.
At least I don't really have that. But I do recommend that you just read as much research as possible from, for example, DoubleThink Lab or DFR Lab or Microsoft Threat Intelligence. You also have Google is also doing research in all of this. There's so many parties out there that have researched spamouflage before. And yeah, you can kind of imitate or copy their research methods. And as a reminder, you can always ask for support and
help within this survey as well. lots of channels for that, which I'm sure the mods can point you to. Particularly when it comes to covering China specifically, we have region -specific channels as well where you can debate and talk about the current news and current investigations. We've also had a recommendation by somebody to check out Sam Cooper's work. We'll go on Chinese interference. There's a Canadian journalist who's uncovered quite a bit about Chinese influence and interference
in Canada specifically. Do you, Jasper, do you have any other recommendations for investigative journalists perhaps that our team, our audience could look at? I'm sorry, what do you mean precisely? In regards to journalists within the space as well, as you mentioned the organizations that are working within the space or any journalists that also stand out to you that are working within this space and highlighting Chinese influence operations. Yeah, so I've worked very closely
together with Wenhao Ma from VOA. So, yeah, he has a sub stack and I'm sure he will lend a nice job again soon and he will keep reporting on it. But I think his China reporting has been really good and he very often also touches on issues related to influence operations. So I definitely recommend following him. And then there is some other, it's not really journalists, but it's more like researchers also. So it's called discourse power. It's also this sub stack.
And they also write a lot of excellent stuff that also touches a lot on PRC influence operations, but also PRC propaganda in general. Amazing, getting so many reading recommendations right now. Thank you. I've put both of those within the chat. We only have five minutes left, so I just want to make sure that we've covered everyone's questions. We had a question earlier particularly about the influence of Chinese operations or
digital operations in Africa. Are you familiar with any large scale influence campaigns by the PRC, for example, on the African continent? And does that differ to their attempts, for example, in Taiwan or in America? So far, I haven't seen, I haven't really been tracking a lot of Africa focused influence operations yet. I'm sure that they They do exist, but in the account pool that we've been tracking over the last few months,
Africa rarely popped up. I would say that countries like the Philippines and Japan are a bigger topic. Also now, obviously, India and Pakistan, that's a very hot topic right now for these accounts. As an African, we start to see a lot of Chinese projects here, mostly in France. Fabulous. Great. Well, thank you so much, Jasper, for your time today. I wanted to just wrap up with a final, final question, if that's okay. What tools are absolutely vital to your deep dive research in
general? You mentioned a couple of threads earlier, but which ones are absolutely vital to them? research that you do. We like to leave people with some kind of key resources to look into. But other than that, thank you very much for joining. And yeah, thank you everybody for listening. Yeah, if you can just throw some tools away and then we'll wrap. Yeah, so I really my single
most important tool is single file. If you don't have access to official APIs or you don't have access to any scrapers, you can use single file. And if you have a little bit of Python or a chat GPT account, then you can figure it out and you can basically get any data and put it into a neat data format and then analyze it. So yeah, definitely just single file. I think we're going to have to start a little Python learning group.
I know quite a few people know it, but I think we should start a little one -up in the Discord server and work together on that. Thanks so much for sharing Single File as a resource for us, and thanks again for joining us on this particular... talk. It's been super fascinating and hope to see you back in the Discord server again soon. Thanks so much, Jesper. Thank you for having me. Thank you, everybody. Thank you for listening
to the stage talk. If you'd like to catch a stage tour live where you can ask the guest questions, join the Bellingcat Discord server by visiting www .discord .gg slash Bellingcat. The music you've heard is titled Dawn by Newer Self and is courtesy of Artlist.