OpenAI Banned Accounts Running Covert Campaigns
OpenAI banned Russian and Iranian accounts that used artificial intelligence to run covert influence campaigns across multiple countries, the company announced on Thursday.
Russian Operatives Target Latin America with Fake Think Tanks
Russian operatives utilized ChatGPT mostly to write internal reports detailing their work to unspecified higher-ups. OpenAI investigated the campaigns and discovered a Latin American operation that created a fake think tank called the Social Research Center. This entity recruited unwitting people to interview experts and compose reports designed to undermine Ukraine’s reputation in the region and influence local politics. OpenAI stated that evidence indicates the Social Research Center’s employees in Latin America did not know they were working for a Russian group and that the majority of original content was produced by this co-opted staff.
In one instance, operatives impersonated a regional educational authority in Peru to trick schools into organizing events featuring Stepan Bandera, a controversial Ukrainian nationalist figure. Bandera is admired by some nationalists for leading a drive for Ukrainian independence, but he is reviled in Poland because his movement killed Jews and Poles and sided with the Nazis. Some schools replied to the fake email address and provided pictures confirming the events took place. A Polish news report about the event prompted a far-right Polish member of the European Parliament to suggest banning supporters of Bandera, successfully feeding into historical tensions between Ukrainians and Poles. The operators also claimed to have planted stories in Poland and Peru alleging that Ukraine was “exporting” ultra-nationalist ideologies.
OpenAI noted this effort was similar to a 2020 operation by individuals associated with the Russian Internet Research Agency, which ran a fake news outlet called “PeaceData” that co-opted journalists globally. That agency was also involved in efforts to influence the 2016 U.S. election.
Iranian Operatives Plant Articles in US and International Media
Iranian operatives took a different approach, using AI tools to help publish approximately 100 articles across roughly a dozen online outlets globally. The campaign focused heavily on the U.S.-Iran war and American politics. Operatives created seven fake journalist personas—such as Ervin B. Hoskins, who claimed to be “an American freelance writer”—to pitch long-form articles to small and medium-sized publications. Hoskins maintained accounts on Meta-owned Instagram and Elon Musk’s X; Instagram data showed the account was based in Iran, while X data linked it to a “West Asia android app.”
Articles with provocative titles like “Is America’s Iran Policy Being Privatized? The Jared Kushner Question” appeared in publications including the LA Progressive and the London-based Middle East Monitor. Other titles included “The Iran War and the Collapse of Western Moral Credibility.” After publication, the Iranian operators asked ChatGPT to generate social media comments to make the stories appear organically discussed by the public, portraying the U.S. as an aggressor and Iran as a resilient victim. These comments targeted topics including diplomatic maneuvering, ceasefires, and missile strikes.
The operation produced articles between July 2025 and October 2026, with frequency increasing after the U.S.-Iran war began. Some personas, including Jenny Williams, Alice Johnson, and Ericka Feusier, listed X handles in their bios that were suspended by August 2026.
Comparative Reach of AI-Driven Propaganda Networks
OpenAI’s investigation highlights a contrast in how actors deploy artificial intelligence. While Russian operatives used AI primarily as a shortcut to update their bosses on field operations, Iranian actors utilized the technology to author convincing prose that bypassed editorial checks at legitimate news outlets. Linvill noted that AI helped Iranian operatives create language that felt authentic enough to be published by real outlets—a hurdle that similar operations would have struggled with years prior. Jessica Brandt, who led the Office of the Director of National Intelligence’s Foreign Malign Influence Center from 2023 to 2025, said it was “remarkable” that Iranians landed content in real media, allowing them to use authentic channels to launder content and expand reach.
AI is also being used in other ways during the conflict; young Iranians have created viral Lego-themed videos to mock U.S. leaders, and President Donald Trump has shared AI-generated videos of U.S. strikes on Iran.
Operatives Used ChatGPT to Draft Reports and Generate Comments
What specific AI tools were used in these influence operations?
The operatives used OpenAI’s ChatGPT to draft reports, translate text into Spanish, and generate social media comments and editorial pitches in Persian and English.
How large was the Russian operation compared to previous campaigns?
OpenAI assessed the Russia-origin operation as a Category 5 on its Influence Operations Breakout Scale, noting it was the most complex attempt to run a front identity that the company has disrupted over the past two and a half years.
Keep reading