OpenAI said on Thursday, Oct. 8, that it had banned two covert influence operations, one from Russia and one from Iran, that combined ChatGPT with older deception tactics to build “false front” entities and get their content into real media outlets. The company detailed both cases in a threat report published that day.
Both operations broke out further than most of what OpenAI has tracked. On the IO Breakout Scale, which runs from 1 (lowest) to 6 (highest), OpenAI placed the Russian operation in Category 5, the first Category 5 case it has disrupted since it began publishing these reports about two and a half years ago. The Iranian operation reached Category 4.
Russia: a think tank staffed by people who didn’t know
OpenAI nicknamed the Russian operation “Dark Clark,” after “Mia Clark,” a fake persona the operators used to run a self-described research platform called the Social Research Center (SRC). Much of the activity aimed to damage Ukraine’s reputation in Latin America, and some targeted local political outcomes, especially in Argentina and Bolivia, according to the report.
The operators’ own reports discussed pay scales, hiring and firing at the SRC, which led OpenAI to conclude they controlled it. “The available evidence indicates that the SRC’s employees in Latin America were not aware that they were working for a Russian group,” OpenAI wrote. Those staffers produced most of the more than 60 articles OpenAI found on the SRC website. “This stands out as the most complex attempt to run a front identity that we’ve disrupted over the past two and a half years,” the company said.
Most of the operators’ ChatGPT use went into drafting internal reports to an unknown superior. A smaller share helped create content, including fake “leaked” documents and audio scripts, some of which OpenAI found circulating online. Several of the fakes drew fact checks and official denials, including a fake audio clip about water shutoffs in La Paz during Bolivia’s May protests, which the state water company EPSAS denied. OpenAI also cautioned that the operators often took credit for events they had nothing to do with, so their internal claims of impact can’t be taken at face value.
The operators asked OpenAI’s models about “Politology,” a reported successor to the Wagner Group, far more than about any other Russian network, CyberScoop reported. OpenAI did not attribute the activity to a specific Russian government agency. Because OpenAI doesn’t allow access from Russia, the operators connected through VPNs.
Iran: seven fake bylines
OpenAI called the Iranian operation “Bogus Bylines.” Its operators prompted in Persian, wrote in Persian and English, and used seven fake journalist personas: Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson. Most posed as American freelance writers covering international politics and the Middle East.
The operators used ChatGPT to check draft articles against specific outlets’ submission guidelines, then to write pitch emails to editors. OpenAI identified almost 100 articles published or syndicated under those bylines across roughly a dozen small and medium online outlets, mostly about the U.S.-Iran conflict, with the earliest from July 2025 and the latest from October 2026. One outlet that ran the articles had almost 2 million Facebook followers as of August 2026, the report said.
Some personas had social accounts whose platform transparency settings gave them away. The Ervin B. Hoskins persona claimed to be an American freelancer, but its X account connected through the West Asia Android app and its Instagram account showed a location in Iran. The operation’s batches of social media comments did much worse than its articles, drawing likes and replies only in the single or double digits, CNA reported. OpenAI said the activity looked like a commercial, for-hire campaign but could not identify the actor.
Old playbook, faster tools
OpenAI said both operations “closely resembled complex influence operations of the pre-AI age,” using AI to speed up parts of the work. It compared the Iranian personas to “Alice Donovan,” a fake journalist used as a front for Russian military intelligence whose articles appeared in Western outlets in 2016-17. Across the 30 covert influence operations it has exposed since early 2024, OpenAI found that those placing content in real media outlets, rather than relying on fake social media accounts, tend to reach the most people.
The company argued that exposure is the best counter. Both “Alice Donovan” and the 2020 fake outlet “PeaceData” stopped operating after they were exposed, OpenAI noted, and it said it has shared information on both new cases with the relevant authorities. In mid-October, X said it had suspended more than 13,000 accounts tied to suspected influence operations linked to Iran, Russia and China, as WebProNews reported.
WebProNews is an iEntry Publication