OpenAI said on Thursday, October 8, that it banned clusters of accounts traced to Russia and Iran for using its ChatGPT assistant to support covert influence operations, CNA reports. In a report, the company said both networks paired AI with older deceptive techniques to run what it called “false front” operations, and used them to “launder geopolitical, conflict-related messaging into their target audiences,” according to CNA. OpenAI did not disclose how many accounts it removed or identify the ultimate operators, CNA adds.

The Russian network, which OpenAI calls “Dark Clark,” focused on Latin American politics and culture, CyberScoop reports. The name comes from Mia Clark, a fake AI persona that the group presented as the leader of the Social Research Center, a supposed regional think tank. According to CNA, the operators’ own ChatGPT-written internal reports said they recruited local people to interview regional experts and write reports for the center’s website. OpenAI said the available evidence suggests those on-the-ground employees did not know that they worked for a Russian operation, CNA reports (paraphrased). CyberScoop quotes a differently worded version of the same point from OpenAI, so this desk treats the exact phrasing as unsettled.

OpenAI described the center as “the most complex attempt to run a front identity that we’ve disrupted over the past two and a half years,” CyberScoop and CNA report. Based on internal reports that discussed wage scales and hiring and firing decisions, OpenAI researchers concluded that the Russian group likely controlled the center rather than merely observing it, CyberScoop writes. NPR reports that OpenAI likened the effort to a 2020 Russian operation that posed as an independent news outlet tied to the late Yevgeny Prigozhin’s Internet Research Agency.

The campaign mainly aimed to damage Ukraine’s image in Latin America, but it also reached into politics in Argentina and Bolivia, according to CyberScoop. In March, the fact-checker Lupa Media flagged a TikTok video that falsely claimed Ecuador’s government used shell companies to recruit citizens to fight for Ukraine, and the actors also spread fake audio of Ukraine’s consul in Ecuador insulting Ecuadorians, CyberScoop reports. During anti-government protests in Bolivia in May, the group released a fake audio clip in which a supposed state water company employee claimed the government planned an emergency and a water shutoff in La Paz, prompting an official denial, according to CyberScoop.

One episode crossed an ocean. NPR reports that the operators said they impersonated a regional education authority in Peru and persuaded some schools to hold activities about Ukraine featuring Stepan Bandera, a Ukrainian nationalist leader admired by some and reviled by others, including in Poland, because his movement sided with the Nazis and killed Poles and Jews. The Peruvian outlet Extra reported that one such event took place, and a Polish news report on it prompted a far-right Polish member of the European Parliament to suggest barring entry to Bandera supporters, NPR says. OpenAI wrote, “The Russian fake thus both fed on, and fed into, historical tensions between Ukrainians and Poles,” according to NPR.

OpenAI rated the Russian campaign a 5 on an impact scale that runs from 1 to 6, CyberScoop reports, and CNA calls it the first Category 5 operation OpenAI has disrupted since it began publishing these reports. CNA ties the rating to evidence that the campaign prompted public comment by politicians in a number of countries; NPR describes one case of a response from a member of the European Parliament. This desk could not reconcile how many politicians responded. OpenAI also cautioned that it could not corroborate all of the operators’ claims, and that they sometimes asked its models to find incidents in target countries that they could take credit for despite having no involvement, CNA reports (paraphrased). The report attributes the activity to no specific Russian agency, but OpenAI said the operators asked about Politology, which CyberScoop describes as a successor to the Wagner Group, “far more” than about any other Russian organization.

The Iranian strand, which The Hill reports OpenAI dubbed “Bogus Bylines,” used ChatGPT to polish articles about the U.S.-Iran war and draft email pitches to real outlets, CNA reports. OpenAI said it found “almost 100 articles published or syndicated under the operation’s bylines across roughly a dozen online outlets around the world,” mostly small and medium outlets focused on international affairs and the Middle East, CyberScoop reports. CNA says the pieces ran under seven fictitious bylines between July 2025 and October 2026, using personas of American or Western journalists, and that platform transparency settings placed at least one persona’s accounts in Iran.

NPR reports that the fake authors placed work in outlets including Daily Kos and the U.K.-based Middle East Monitor, and that, according to NBC News, Middle East Monitor published at least 21 articles by four of the fake authors and has since removed the author pages. Daily Kos told NPR that the persona posted in its community section, which staff do not review before publication; those posts remain up, NPR says. OpenAI rated the article-placement strand Category 4, CNA reports. A separate strand that generated English and Persian social media comments drew likes and replies only in the single or double digits, many from the operation’s own accounts, and OpenAI said the operators’ internal reports “used a deceptive calculation” to measure that impact, according to CNA.

Researchers see a lesson in how the operatives got caught. Darren Linvill of Clemson University’s Media Forensics Hub told NPR that the Russian operatives used AI to “create a shortcut and do less work,” and that “it helped get them caught.” On the Iranian side, he said AI “helped them create language that … looked and felt and read authentic,” NPR reports. OpenAI has published similar reports before; in June 2026 it said it disrupted two Chinese-linked operations, CNA notes.