How Venezuela's government uses AI as propaganda - Trolled podcast, BBC World Service

For me it's a hobby. I don't work for the government but I've always liked Twitter. I was one of the first ones to join in 2009. My first account was blocked a year ago. I had 14,000 followers. You're hearing from a digital propagandist for the state of Venezuela. We'll call him Rafael, that's not his real name. The aim is to amplify the information that the government puts on Twitter. He's a propagandist not because he thinks life is great in Venezuela but because it's so tough for him. There is a lot of reservations because, well, too much is lacking. They haven't fulfilled their job for the people as they should have. Venezuela, we find, has connected their social welfare system to Twitter. To reward people like Rafael who most loyally pump out pro-government hashtags. This is BBC Trending's Trolled, the series lifting the lid on online abuse and the forces at play on the dark side of social media. I'm Carl Miller, an expert in how states and others try to secretly influence the information that we see. Today I'm going to tell you about a strange ecosystem of digital control. It's one which connects Venezuela's social welfare system to many pro-government hashtags people pump out. It's a story that involves thousands of real people and experiments to replace people and it stretches from Rafael all the way up to the Venezuelan president Nicolas Maduro himself. Our story begins with this news report. Listen carefully. 2023 Caribbean series was organized this year in Venezuela and to the surprise of many it was a total success. It first appeared on a mysterious YouTube channel called House of News which broadcasted hyper-localized Venezuelan news in English before it was closed down. Can you hear? The reporter sounds stilted. Wooden. Media from different countries have described the event organized by Venezuela as a total and unexpected success. That's because it isn't real. Nor is the report. Artificial intelligence was used to generate something that very nearly looks like a journalist. It's called a deepfake. The idea of this one and a number of others was to show Venezuelans that the West was talking positively about their country. That it had international prestige. No one knows who actually made them but from an almost unknown YouTube channel they made it onto state television accompanied by Spanish subtitles. Then they were exposed as deepfakes and all hell broke loose. Adrián González runs one of the most well-known anti-disinformation websites in Venezuela called Cazadores de Fake News. He and his team instantly recognized them as a campaign to promote a pro-government Venezuelan narrative. When we realized that it was a paid video we immediately made the connection that other channels like this one used to exist. They also had very few followers but thousands of views precisely because they were paying for publicity on YouTube and that they had other characteristics. In some cases videos that were of caricatures or satire but all of it was publicity that we call cover propaganda or disinformation with the same narrative as House of News. But House of News took their campaign a step further than Adrián had seen before. They had a very specific characteristic. It was the type of fake news that used presenters but they weren't presenters. They were avatars made with artificial intelligence with a program called Synthesia. Synthesia, a tech company based in London, uses AI to generate videos. For this particular campaign Adrián found it was their technology that was used to make it look like Western journalists were lauding the achievements of the Venezuelan state. We asked Synthesia whether they could track down who was behind these deepfakes and what they're doing to stop their services from being used maliciously. They didn't answer our specific questions but they told the BBC they're involved in building a safe future for their industry and that moderation and regulation are key to ensuring the responsible development of AI. There were videos mainly about the supposed improvement of Venezuela's economy. There was another about tourists that had visited Venezuela during Carnival and the potential it reflected as a tourist destination. Another video about Easter. Another creating a narrative that was against the opposition. They are all corrupt. They are chief. They have little credibility. Advances in AI are providing fertile ground for governments accustomed to spinning stories and opening up new opportunities for those trying to deceive us. Here in London at the BBC's headquarters I met with Henry Ida, an expert in synthetic media, and I played in the deepfakes we've been looking at. As the synthetic world like melts into the real one, how worrying is all of that for us? When those kinds of images can fool people that easily it is a cause for concern. Obviously in more high-stakes scenarios, where perhaps there's a lot riding on whether it's real or fake, people might be more critical and more reflective about the content, maybe look for some kind of signs that it might be it might be fake, but that's not going to be sustainable for the future because the signs that we look for now are not necessarily even going to be there tomorrow. So it is getting really difficult to spot the content and it's only going to get more difficult particularly as video and text to video gets better. Tens of thousands of people watch these deepfake videos online. Many would have thought they were real news reports. But as an expert in disinformation, Adrian, like Henry, knew the signs. In a tell-all Twitter thread, Adrian exposed them for what they were. It didn't gain much traction until the prestigious Spanish newspaper El País picked up the story. The government's response was swift and emphatic. Freddy Néñez, the Minister of Communications, used Twitter to denounce the paper's accusations as baseless, calling their reporting a conspiracy theory and emphasized the success of the revolution was due to a more superior threat. It wasn't artificial intelligence, but the country's social and popular intelligence. Every day in Venezuela, his ministry announces a hashtag of the day, a mixture of slogans and phrases and ideas, all pro-government of course, parroted by the government and its allies online, in the media and at rallies. It announced the hashtag of the day was Somos Inteligencia Social. We are social intelligence. Now, that might mean nothing to you. It certainly doesn't mean much to me. But if you live in Venezuela, whether it's in context or not, those words mean a lot and we'll explore that later in the episode. The government kept going and Minister Néñez shared a different deepfake video, also made using Synthesia's technology. The avatar in the video is saying, hi, what do you think we should do to strengthen the revolution? It doesn't directly address the allegations made in El País's article, but it's clear it was made to poke fun. We're going to pause here for a minute. There is a lot going on. We've got deepfakes pretending to be Western journalists saying nice things about Venezuela. We've got the revelation of these deepfakes from a real Western outlet. And we've got a government who will not confirm or deny they were behind it. And in the background of all of that, we have Somos Inteligencia Social, we are social intelligence, being promoted by the state. And amidst all of this tangle, President Nicolas Maduro himself intervenes. In a televised speech on state media, he spoke about the newspaper's coverage of Venezuela coming from a place of European arrogance, jealousy, ego and hate. That it's not artificial intelligence Venezuela is using, it's popular intelligence, a revolutionary intelligence, something the West cannot comprehend. The president singled out El País, the newspaper that exposed the deepfakes. But it was Flor Antonia Singer, the Venezuelan journalist who wrote the article for the newspaper, who bore the brunt of criticism from his supporters. How bad did it get? Do you know whether she fears for her safety? Well, I'm sure that if she's not speaking about this as she regularly will do, she's afraid of the consequences this could have in the future. Flor Antonia wouldn't speak to us about the abuse she received. But Marivi Vazquez, executive director of Provox, an NGO combating online disinformation in Latin America, worked with her on the story. And she told us that Flor Antonia had become the latest target of a trolling frenzy by the government and its puppets, which was taking place online but also offline. This isn't unusual in Venezuela. Reporters Without Borders ranks the country as having one of the least free media environments in the world. And the Venezuelan government has a track record of quashing any form of dissent. Yet even against this backdrop, Flor Antonia has tackled some difficult, hard-hitting subjects. Corruption, human rights abuses, much of her reporting holding President Maduro's government directly to account. So why did this particular story catch the attention of Maduro and his government? As it stands, there is no concrete evidence that these deepfakes are actually part of a government propaganda project. But get this. It was shown on VTV, a state-owned channel whose president is none other than our friend, the Minister of Communications, Freddy Nanez. Adrian and other experts we've spoken to think it's safe to assume that there is some connection and that the government went into damage control. There is a need to control the narrative in the country. But what they cannot control is forced by propaganda and used in some way to try to neutralise information that can be disadvantageous for the government. One of the oldest tricks up their sleeves, and possibly the most successful in their history, is the use of revolutionary rhetoric to gain public support. Remember those words? Social intelligence, popular intelligence, revolutionary intelligence? They aren't random. They mean something if you're Venezuelan. BBC Monitoring's Venezuelan expert, Rochelle Krieger, explains their significance. These words are grounded in chavismo, a left-wing populist ideology formed under President Maduro's predecessor, Hugo Chavez. Chavez, who is almost deified in Venezuela, championed a nationalistic, anti-West discourse. He sold this idea of freedom from the West, its oppression and its capitalism. The rhetoric being used here is exactly that. It's textbook chavismo. The government is using the allegations of artificial intelligence to flip the narrative, saying that the people of Venezuela are all in it together and have the power to decide what they do, want and believe. We put the question of whether the Venezuelan government is actively using deepfake technology to spread propaganda to their Ministry of Communication. We also asked them why they broadcast deepfake disinformation on state news, but they did not respond. It's a strange story, and this is the strangest connection. Whilst House of News was experimenting with AI to create fake people, the government's response relied on thousands of real people. And one of them was Rafael, the online propagandist who we heard from earlier. You have to do it in a post way to avoid being blocked. So I do like 100 in the morning and 100 in the afternoon, for example. He's talking about the number of tweets he pushes out in favour of the government in a single day. Let me explain how it works. These hashtags are first tweeted from the official Ministry of Communication's Twitter account. And then once the Ministry tweets out a hashtag, it's amplified by people like Rafael who are paid to share propaganda, as well as some elected officials' accounts and their sympathisers. Marie Vivasquez from Provox explains. The use of this is basically to amplify any narrative they want online. And this makes it harder to understand what's happening in the country because you have all this pollution of poll messages. The purpose is to game the algorithm and occupy the trending box with your own slogans, even if they are completely meaningless. And this is very dangerous because when you go to see the trending topics and you see that most of the hashtags are supposedly supporting the government, maybe you can get confused about what is actually happening in the country. Hashtag SomosInteligenciaSocial, we are social intelligence, served a specific purpose. It drowned out discussion of the deepfake. Provox examined the use of the hashtag from the first time it appeared on Twitter. At the time of recording, they found it was used in over 360,000 tweets. Sent by how many people? Well, that's hard to tell. More than 90% of the accounts interacting with the hashtag, according to our studies, were accounts with unauthentic behaviour. This means they could be bots or they could be digital troops being paid to amplify these false narratives online. Provox says it's both bots and digital troops, but what they're unable to clarify is how many digital troops there actually are. So what motivates Rafael to tweet propaganda? Well, it's not his love of the current government. I was a fan of Hugo Chavez, but of Maduro, not really. I only support him a bit. A lot of people say that Venezuela is an oil country, but they don't know how to manage their resources. For Rafael, the money is worth it. Venezuela has seen an economic collapse over the past decade and inflation and shortages have become just part of life. So the extra 10 US dollars or so that Rafael makes, on top of his monthly salary of around 80, gives him the freedom to buy more of the things he really needs. Look mostly to help myself with the payments. The two bonuses a month, like 242 bolivares. With that, I can buy, for example, three cups of flour and some oil and rice to cover a little bit of the food. The whole system is gamified through Sistema Patria, a government-owned and administered social security app used by most Venezuelans across the country, which gives them access to subsidies and allows them to receive public salaries, state pensions, COVID-19 vaccines, but also doubles up as a platform for paying digital troops like Rafael. We have no access to the government at all. Through the Ministry of Communications' Twitter account, who publish the hashtag, each Twitter copies the hashtag of the day and either retweets it or writes up their own personal tweets with the hashtag. According to experts who have studied payments through the app, people like Rafael have to synchronise their Twitter account with Sistema Patria. This gives Sistema Patria access to Rafael and countless others' data to monitor their activity on the app. And depending on what you're tweeting, every month digital troops receive what's called a bonus in their digital wallet. The minimum you can receive is about one and a half US dollars. The more you tweet, the more money you make. First place is about 90 bolivares. Second, I think, is about 60. Third place is about 52. And fourth is 42 and a half bolivares. The results are announced on Mondays on the account of the Patria platform. The amount Rafael receives ranges from about two to three and a half dollars a week. We've seen some of the payments Rafael's received this past year in screenshots he sent us. They are described as social media mention awards. It doesn't sound like a big deal, but... Glue hundreds of thousands of Rafael's together and you end up with a form of social control, where the government can drown out online criticism and always keep the trending box full of pro-regime slogans. We put to the government all that we found out, that they are paying Venezuelan citizens to tweet propaganda on behalf of the state. We also informed them we had spoken to a tweeter who had shown us statements with payments from the government and asked them to explain why they're doing it. They did not respond. What we're witnessing here is not just a new era of Venezuelan propaganda tactics, but it's maybe something that actually we're all having to navigate. A world where our distinctions of human and bot, real and artificial, well, they're all coming under more and more pressure. In this story, we've got the creation of fake journalists and then the revelation of these fake journalists by the real journalists and real people behaving like bots on behalf of the government. And whilst much isn't clear, one thing I think is, things won't stand still. Venezuela is experimenting even now with new forms of technology to control the narrative. Last month on VTV, the same state news channel that aired these deepfake reports, President Maduro introduced CIRA, an artificially made female news presenter on his new TV show, Con Maduro. Some experts have suggested that this may be a jab at the recent accusations of government propaganda campaigns. But Venezuela says it's advancing in its use of technology and AI. That's it for this edition of BBC Trending's Trolled with me, Carl Miller. Thank you so much for listening. BBC Trending will be back with a new series later in the year. A huge thanks to producers Reija Canzara and Rochelle Krieger, editor Flora Carmichael and Annie Gardiner for mixing this episode. Trending is also on Facebook and Twitter just like this. And remember to leave us a rating or review wherever you get your podcasts from.