It was the kind of discovery that sounds like a spy novel but is now a routine reality in the digital age. On Thursday, October 8, OpenAI—the company behind ChatGPT—published a report revealing that agents linked to the Iranian regime had been quietly using its advanced AI tools to produce fake news articles designed to shape public opinion against the United States. The articles were not just sitting on obscure websites; they were successfully pitched to and published by real international news outlets, complete with fabricated Western bylines. According to the report, these operatives created seven entirely fictional journalist personas, gave them believable names and backgrounds, and then used those personas to shop AI-written stories to editors. The operation ran from July to September, and during that window, OpenAI investigators identified nearly 100 articles that had appeared under these fake names across a dozen small and medium-sized news organizations. It was a coordinated, low-key but determined effort to inject Iranian talking points into the global conversation—and for a time, it worked.
The mechanics of the operation are both impressive and unsettling. The Iranian agents used OpenAI’s ChatGPT to draft entire articles, then attached them to fake journalist identities designed to look credible to editors. These personas likely had fabricated social media profiles, professional biographies, and contact information, making them hard to distinguish from real freelance writers. The articles were not random noise; they were tailored to fit the editorial style of each outlet and written in ways that mirrored Western journalism conventions. To make the posts appear popular and legitimate, the operation also generated positive social media comments, boosting engagement and giving the illusion that these pieces were being read, shared, and appreciated by real audiences. When OpenAI detected the scheme, it moved quickly to block the ChatGPT accounts associated with the activity. But by then, the damage was already done—dozens of articles had already been published, and the false narratives had entered the public domain, where they could be shared, cited, and absorbed before any corrections could be made.
The headlines OpenAI flagged reveal the clear political direction of the campaign. One article was titled “Trump’s Iran Gamble Hardened the Regime it was Meant to Break,” suggesting that U.S. pressure only strengthened Tehran. Another asked, “Is America’s Iran Policy Being Privatized? A Question for Jared Kushner,” implying that American foreign policy was being driven by personal business interests rather than national security. A third declared, “Why Trump Blinked on Iran,” framing the former president as weak and indecisive in the face of Iranian resistance. These are not subtle messages. They all push the same narrative: the United States is incompetent, divided, or corrupted, and Iran is the wronged party simply defending itself. The use of AI allowed the Iranian operatives to generate a high volume of such content quickly, adapting their messaging to current events and pitching it to journalists who were likely unaware they were being used as conduits for foreign propaganda. The stories were designed to look like independent journalism, but they were, in effect, paid political advertisements for the Iranian regime, disguised as neutral reporting.
This incident is the latest and most sophisticated example of AI tools being weaponized in the ongoing propaganda war between Iran and its adversaries. The backdrop is important: in February, the United States and Israel launched a joint offensive aimed at stopping Iran’s nuclear weapons program and other activities that Washington and its allies view as threats. Since then, the information space has become a battlefield in its own right. The Iranian regime has long used media outlets, social media accounts, and even front organizations to spread its message, but the advent of generative AI has dramatically lowered the cost and increased the scale of such operations. Instead of hiring writers, translators, and social media managers, a small team of operatives can now generate hundreds of articles in multiple languages with a single tool. AI can imitate the tone of a seasoned political analyst, cite credible-sounding sources, and even produce variations of the same argument for different audiences. The result is a propaganda machine that is faster, cheaper, and harder to trace than anything seen before. OpenAI’s report is a reminder that the same technology that powers chatbots, translation tools, and creative assistants can also be used to manipulate public opinion, undermine trust in journalism, and blur the line between authentic reporting and state-sponsored messaging.
The news outlets that unknowingly published these AI-generated articles were not mainstream giants like The New York Times or CNN, but that does not mean they were insignificant. The twelve outlets identified by OpenAI are described as small to medium-sized, yet some command substantial audiences. For example, the London-based Middle East Monitor, which focuses on news related to Palestine and the broader Middle East, has nearly two million followers on Facebook. That is a significant reach, especially for a niche publication, and it means the fake articles likely reached far more people than the raw number of outlets might suggest. When CNN reached out to Middle East Monitor’s editor-in-chief, Dawoud Abdullah, he said he would review the articles flagged by OpenAI and “take appropriate action.” A representative later confirmed that the articles had been removed and that the publication was reviewing its submissions process to prevent similar incidents in the future. That response is encouraging, but it also underscores a broader vulnerability: many smaller news organizations rely heavily on freelance submissions and do not have the resources to thoroughly vet every writer who pitches them. In an environment where AI can generate convincing articles in seconds, even a well-intentioned editor can be fooled by a fabricated persona and a polished pitch.
What does this mean for the average reader? It means the information we consume online is becoming increasingly difficult to trust. We already live in a world of deepfakes, manipulated images, and viral misinformation, but the Iranian operation shows that AI is now being used to infiltrate the very institutions we rely on to separate fact from fiction. The articles published under fake bylines were not obviously false in the way that a hoax or a conspiracy theory might be. They were carefully crafted opinion pieces that advanced a particular political agenda while appearing to be legitimate journalism. This is far more insidious than a fake social media post because it exploits the credibility of established media brands. When a reader sees an article on a news website, they assume it has been edited, fact-checked, and approved by professionals. But in this case, that trust was betrayed—not by the editors, but by the invisible hand of a foreign intelligence operation using cutting-edge technology to deceive them. OpenAI’s decision to block the accounts and publish its findings is an important step, but it is only a partial solution. The AI tools themselves are neutral; they can be used for good or ill. The real challenge lies in the hands of journalists, platform moderators, and readers, who must become more skeptical, more vigilant, and more willing to question the origin and authenticity of the content they encounter. As this story from VOA Persian shows, the battle for truth in the age of artificial intelligence is just beginning—and it will require not just technical solutions, but a renewed commitment to transparency, accountability, and critical thinking across every level of the global information ecosystem.