The Pulse
OpenAI Disrupts Russian and Iranian Influence Operations
OpenAI says it banned Russian and Iranian ChatGPT account clusters tied to separate influence operations. Its investigation identified nearly 100 articles published or syndicated under seven fake journalist bylines linked to the Iran-origin
AI.info Team · Written with AI ·
Nearly 100 articles appeared under seven invented journalist bylines in an Iran-origin influence operation, OpenAI said on October 8, 2026, as it announced bans on ChatGPT accounts tied to separate Russian and Iranian campaigns. The articles reached more than a dozen online outlets, while the Russian operation built a Latin American research group that OpenAI says was run by Russian actors through a false identity. Both campaigns used AI alongside conventional influence tactics, but their apparent reach differed sharply.
Russia’s “Dark Clark” built a front in Latin America
OpenAI says it banned a cluster of accounts originating in Russia that used ChatGPT to draft internal reports, create campaign material and prepare assessments of a self-described research group called the Social Research Center. The operators appeared to control the group through a fake persona, “Mia Clark,” while people working on the ground in Latin America were apparently unaware of the Russian connection. The company says the group published more than 60 articles, most of them apparently original work.
Campaign material targeted Ukraine and local political debates, particularly in Bolivia and Argentina. OpenAI found examples including false audio and videos about alleged Ukrainian recruitment in Ecuador, and a fabricated recording claiming Bolivia’s water company would shut off service in La Paz. OpenAI says open-source research corroborated some fakes through fact checks and official denials, but it could not verify others; the operators also claimed credit for events their campaign did not cause.
OpenAI rated the Russian operation Category 5 on the six-level IO Breakout Scale, its highest rating for an operation it has disrupted. The company points to apparent public responses to some fabricated stories, including political comment in several countries, while cautioning that the operators’ own impact reports were unreliable.
“Bogus Bylines” placed articles in online outlets
The Iran-origin operation used seven invented author identities to pitch English-language articles about the US-Iran conflict. OpenAI says its investigation found almost 100 articles published or syndicated across roughly a dozen outlets, with publication activity increasing after the conflict began. Some outlets promoted the articles through their own social accounts, extending their potential audience beyond the websites themselves.
The operators also asked ChatGPT to generate batches of comments in Persian and English, often in response to posts about the conflict involving Iran, the United States and Israel. Some comments praised or endorsed the operation’s own articles. OpenAI found no evidence that its models helped automate the posting, and says the comments it located generally attracted little engagement and formed a minority of replies.
The company assessed the article-publishing part of the Iranian campaign at Category 4, but rated its social-media commenting at Category 2. One outlet that published the operation’s articles had nearly 2 million Facebook followers, about 355,000 on X and more than 544,000 on Instagram as of August 2026, according to OpenAI. Those figures describe the outlet’s audiences, not the number of people who read or engaged with the campaign’s articles.
AI assisted the workflow, not every part of the scheme
OpenAI’s account depicts AI as one tool in broader operations built on fake identities, editorial pitches, social accounts and, in the Russian case, unwitting local workers. The company says the Russian actors used ChatGPT most heavily for internal reporting; requests to create deceptive content were a smaller share of their activity. In the Iranian case, the model helped refine articles and pitches and generate comments, while the operators’ internal reports overstated impact by counting views on original posts rather than views on their replies.
OpenAI says it banned the accounts and shared information about both cases with relevant authorities. It did not identify the commercial actor it believes may have run the Iranian operation, and its public account does not establish the identities of individual operators. The evidence it describes shows a particular risk of false-front campaigns: articles placed with real publishers can travel through audiences those operators could not reach by posting only from fake social accounts.