Imagine sitting down at your desk, opening ChatGPT, and typing: “Draft a report for my superiors about how I successfully spread propaganda in a foreign country.” It sounds absurd, like something from a spy novel with a very bureaucratic villain. But according to OpenAI, that is essentially what happened. On Thursday, the artificial intelligence company announced that it had banned a network of Russian and Iranian accounts that were using its chatbot to help with influence operations across multiple countries. The operatives were not using the technology to build deepfakes or create viral memes. Instead, they were using it for something far more ordinary: writing internal memos, polishing articles, and updating their bosses on how well their schemes were going. The revelations are striking not because they show AI creating brand-new threats from scratch, but because they show how human these operations really are. There are people behind the propaganda, and those people have deadlines, performance reviews, and a need to look productive. They used ChatGPT the way many office workers might use it—to save time, to sound smarter, and to get the job done. The difference is that their job was to deceive people, undermine democracies, and exploit historical wounds for political gain.
The most dramatic example OpenAI uncovered involved a Russian operation targeting Latin America, particularly Peru, with the goal of undermining support for Ukraine and influencing local politics. The operatives impersonated a regional educational authority and tricked schools into organizing activities about Ukraine that featured Stepan Bandera, a deeply controversial figure in Eastern European history. Bandera is admired by Ukrainian nationalists for leading a drive for independence before and during World War II, but he is reviled by others, including many Poles, because his movement sided with the Nazis and was involved in the killing of Poles and Jews. The Russian operatives knew exactly what they were doing. By placing Bandera at the center of school events in Latin America, they were not just spreading a fake story; they were reaching into one of the most painful and complicated chapters of Polish-Ukrainian relations. The plan worked. A Peruvian outlet, Extra, reported that one such event did take place at a Peruvian school. A Polish news report then picked up the story, and a far-right Polish member of the European Parliament responded by suggesting that people who professed support for Bandera should be banned from entering the country. OpenAI noted that the Russian fake “both fed on, and fed into, historical tensions between Ukrainians and Poles.” This is the essence of what experts call narrative laundering: taking a manufactured story, running it through real institutions, and letting it emerge as something that looks like legitimate news. The fact that a political figure in Europe reacted to it shows how a single fake event, amplified by real media, can ripple outward in ways that are impossible to control.
The Russian operatives were not just staging school events. According to OpenAI, they also ran a fake think tank in Latin America and recruited unwitting researchers to interview experts and compose reports. This is a classic influence operation tactic. By creating an institution that sounds credible, operatives can get their ideas published, quoted, and treated as serious scholarship. It is a way of laundering propaganda through the appearance of academic legitimacy. OpenAI noted that the effort was similar to a 2020 Russian operation that posed as an independent news outlet and was linked to the late oligarch Yevgeny Prigozhin’s Internet Research Agency, the same organization involved in efforts to influence the 2016 U.S. election. What is particularly revealing is how ordinary the Russian use of AI was. The operatives used ChatGPT to generate fake material, but that was only a minority of the work. Mostly, they used it to write reports about what they had done, essentially creating a paper trail for their superiors. That turned out to be a mistake. “These Russian underlings were trying to use AI to create a shortcut and do less work in the same way that a lot of us use AI, but in this case it helped get them caught,” said Darren Linvill, a co-director of Clemson University’s Media Forensics Hub, who studies online influence operations. There is something almost darkly funny about the image: a propagandist, hunched over a laptop, asking a chatbot to help summarize his efforts to destabilize a region. But it also reveals a deeper truth about disinformation. It is often not the work of evil geniuses. It is the work of people trying to get through the day, meet expectations, and avoid getting in trouble.
The Iranian operations OpenAI uncovered had a different flavor, but they were just as concerning. While the Russian operatives used AI mainly for internal communication, the Iranian operatives used it to help push out content in Persian and English. The campaign focused primarily on the U.S.-Iran conflict and, to a lesser degree, on American politics. OpenAI said the operation’s activity “appeared consistent with a commercial actor running a for-hire influence campaign,” though the company was unable to identify exactly who was behind it. The operatives created seven fake authors and used them to pitch and polish long articles to small media outlets around the world. In total, they managed to publish nearly one hundred articles this way. Some of those articles appeared in the progressive U.S. news site Daily Kos and the U.K.-based Middle East Monitor. According to NBC News, the Middle East Monitor published at least twenty-one articles by four fake authors. The outlet, which did not respond to a request for comment, has since taken down the author pages. Daily Kos, for its part, said that its community section, where one of the fake personas posted, is separate from staff reports. Anyone can sign up to post in the community section, and posts are not subject to review by Daily Kos staff before publication. The site has rules against pretending to be someone you are not, but the posts by the fake persona were still up at the time of reporting. This is the messy reality of the modern information ecosystem: platforms are struggling to police content, especially when it comes from accounts that look and sound like real people.
The Iranian operation highlights something that experts say is genuinely new about AI in influence campaigns. It is not just a shortcut; it is a tool that can make propaganda feel authentic in ways that were previously very difficult. “AI was valuable to them above and beyond just being a shortcut,” Linvill said. “It helped them create language that looked and felt and read authentic to the point where it was being published by legitimate outlets. And I think that is something that a few years ago, an Iranian operation like this would have struggled a lot more with.” Think about what that means. Before AI, a foreign operative trying to publish an article in an English-language outlet would have to write in a second or third language, making mistakes that would raise red flags. They would need a human editor or translator, which takes time, money, and coordination. AI removes many of those barriers. It can produce fluent, idiomatic prose in seconds. It can imitate the tone of a particular publication. It can generate convincing biographies for fake authors. This is a powerful capability, and it is available to anyone with an internet connection. The result is a media landscape where a reader can no longer assume that an article with a byline was written by a real person, or that a quoted expert is a real scholar, or that a story picked up by multiple outlets is true.
So what should we take away from all this? First, AI did not invent disinformation. Governments and political actors have been spreading lies, manipulating the press, and stoking social divisions for as long as there has been politics. What AI does is make those efforts cheaper, faster, and more convincing. It is a force multiplier, not a magic wand. Second, AI is a double-edged sword for the people who misuse it. The same tools that help operatives write reports and articles also leave trails that researchers and companies can follow. In this case, the Russian operatives’ use of ChatGPT to document their own work is what allowed OpenAI to investigate and ban them. They were caught, in part, because they outsourced their memory to a chatbot. Third, and perhaps most importantly, these operations work because they exploit human trust. A school in Peru does not expect a fake educational authority to send them lesson plans about a controversial Ukrainian nationalist. A small news outlet does not expect a polished, well-sourced article from a nonexistent author. We want to believe that the institutions around us are real, that the news we read is written by actual people, and that the stories we share are true. Disinformation campaigns are designed to take advantage of that trust. The defense is not just better AI detection tools, though those are important. It is also skepticism, media literacy, and a willingness to ask who is behind a message and why they want us to believe it. The people running these operations are not faceless machines. They are operatives with bosses, deadlines, and ChatGPT tabs open. They make mistakes. They get caught. But they will keep trying, because the goal is not to convince everyone of one lie. It is to make it impossible to know what is true.