OpenAI Says a Russia-Origin Operation Ran a Latin American 'Think Tank' Through Apparently Unwitting Staff, and an Iran-Origin One Placed Nearly 100 Articles Under Fake Bylines

OpenAI's newest threat report describes two influence operations that used ChatGPT as much to run the machinery around their propaganda as to polish it: pitching editors, managing a front organisation's staff, and writing flattering reports for their own bosses. In "Disrupting AI-enabled 'false front' operations," published October 8, OpenAI says it banned two clusters of accounts, one originating in Russia and one in Iran, and shared information on both with the relevant authorities.

OpenAI rates the Russia-origin operation Category 5 on the IO Breakout Scale, which runs from 1 (lowest) to 6 (highest), and says it is "the first Category 5 operation we've disrupted since we began our reporting." The Iran-origin operation's article placements reached Category 4. OpenAI says it has exposed 30 covert influence operations in the last two and a half years, and that those which land content in real media outlets tend to have the highest potential reach.

Dark Clark: a front research center in Latin America

According to OpenAI, Russian-speaking operators, using VPNs because OpenAI blocks access from Russia, worked mostly to undermine Ukraine's reputation in Latin America, with some efforts aimed at domestic politics, especially in Argentina and Bolivia.

The core of the operation was a self-described research platform, the Social Research Center (SRC), whose website says it focuses on the Indian diaspora in Latin America. The operators' reports discussed pay scales, hiring and firing, which OpenAI reads as control rather than cooperation. They ran its social media through a fake persona, "Mia Clark," which gives the operation its nickname. OpenAI says "the available evidence indicates that the SRC's employees in Latin America were not aware that they were working for a Russian group." It found well over 60 articles on the SRC website, most of them apparently original, and calls the effort "the most complex attempt to run a front identity that we've disrupted over the past two and a half years."

OpenAI contrasts the SRC with "another Russia-linked 'think tank' that we recently exposed," which plagiarised most of its content. It does not name that operation. In August, we reported on OpenAI's account of a Russia-origin cluster promoting the International Burke Institute, most of whose sampled articles were copied and which OpenAI rated at the low end of Category 3.

Most of the operators' ChatGPT use went into internal reports to "an unknown superior." Those reports claimed a string of fakes. OpenAI says it found matching material in public for several of them:

• Peru: a fake email from Lima's regional education directorate asked schools to hold Ukraine-themed events referencing Stepan Bandera. Stories matching the operators' claims then appeared in the Peruvian and Polish press, and some included a Polish member of the European Parliament proposing that people showing "anti-Polishness" be declared persona non grata.
• Ecuador: a fake email urged schools to pledge allegiance to President Daniel Noboa and to Erik Prince, the Blackwater founder. OpenAI found press coverage matching the claim and a detailed rebuttal from Ecuador's education minister. For a claimed fake audio clip of Ukraine's consul in Ecuador, OpenAI found a matching TikTok post that drew limited engagement.
• Bolivia: fake audio of a worker at the state water company EPSAS claimed the government would cut off water to La Paz during late-May protests. OpenAI points to a debunk and an official denial from EPSAS.

Some fakes had already been tied to "Politology" or "La Compania," which open-source researchers describe as a reported successor to entities founded by Yevgeniy Prigozhin. OpenAI cites that public reporting and notes the operators' interest in Politology, but, as CyberScoop also observes, the report does not attribute the activity to any specific Russian government or intelligence agency. It also warns that the operators took credit for events they had nothing to do with, so "their own claims of impact in their internal reporting cannot be taken at face value." The Category 5 rating rests on fakes OpenAI could corroborate, including evidence that the operation "led to public comment by politicians in a number of countries."

Bogus Bylines: seven fake journalists

The Iran-origin operators prompted in Persian, used VPNs, and asked ChatGPT to check English-language articles about the US-Iran conflict against specific outlets' submission rules, then draft pitch emails to editors. They pitched under seven bylines: Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson, most presented as American journalists. Some personas had backing accounts on Medium, X, Facebook, Instagram or Threads. OpenAI says transparency settings on the Hoskins accounts placed them in Iran.

OpenAI identified almost 100 articles published or syndicated under these bylines between July 2025 and October 2026, in small and medium online outlets focused on international affairs and the Middle East. One of them had almost 2 million Facebook followers as of August 2026. OpenAI's count of outlets differs between passages ("roughly a dozen" in one, "over a dozen" in another). Publication picked up sharply after the US-Iran conflict began.

The operation's batch comments on social media fared worse. OpenAI found engagement in the single or double digits, often from the operation's own accounts, and rates that strand Category 2. The operators' internal reports counted views on the posts they replied to rather than on their replies, a method OpenAI calls deceptive and says "greatly exaggerated" their impact. OpenAI says the activity looked "consistent with a commercial actor running a for-hire influence campaign, but we are unable to identify the particular actor involved."

Limits

• Every finding, attribution and Breakout Scale rating here is OpenAI's own assessment; Times of AI did not independently trace the fakes, outlets or accounts. The report does not say how many accounts were banned, as CNA notes.
• OpenAI's corroboration relies on matching operators' claims to public material; for several claimed fakes it found nothing, possibly because they were removed or never posted.
• The individual outlets that published Bogus Bylines articles are not named in the report text read for this article.

Sources

• OpenAI: Disrupting AI-enabled "false front" operations (October 8, 2026)
• CNA: Russian fake think-tank with real employees exposed in 'most complex' influence operation attempt, OpenAI says (October 9, 2026)
• CyberScoop: OpenAI says Iran, Russia used AI journalists, think tanks to influence Western media (October 8, 2026)
• Times of AI: OpenAI Bans a Russia-Origin Cluster Promoting a Fake Israel Think Tank