OpenAI: Russian 'Dark Clark' Network Is the Widest-Reaching Influence Operation It Has Disrupted

Since OpenAI's Thursday, Oct. 8 report disclosed that it had banned accounts behind an Iranian campaign that placed nearly 100 articles under seven fake bylines, the details of the second operation in that report have come into focus. The Russian effort, which OpenAI calls "Dark Clark," is the one the company says matters more.

OpenAI rates Dark Clark a Category 5 on the six-point IO Breakout Scale. That is the highest rating it has assigned in two and a half years of publishing these reports, the company said. The rating rests on one fact: a politician reacted to the content.

A think tank run by people who did not know

The operators ran the Social Research Center, or SRC. It publicly described itself as a research platform on the Indian diaspora in Latin America and on India's ties with countries in the region. OpenAI says Russian operatives controlled it through a fabricated identity.

The staff on the ground were real. OpenAI said the evidence indicates the SRC's Latin American employees "were not aware that they were working for a Russian group" and did their research "in good faith." The SRC appears to have produced a majority of its content through those co-opted staff, not through AI. OpenAI called it "the most complex attempt to run a front identity that we've disrupted over the past two and a half years."

The goal was to damage Ukraine's reputation in the region. Some activity also appeared aimed at domestic politics in countries including Argentina and Bolivia.

Fake documents, fake school notices

The operators used ChatGPT to create fake "leaked" documents and scripts for audio recordings. Some circulated online. At least some of the false claims spread widely enough to prompt fact-checks and official denials, including in Ecuador and Peru, according to OpenAI.

The most consequential episode, OpenAI said, was described in the operators' own internal reports. They claimed to have impersonated a regional educational authority in Peru and to have tricked some schools into organizing activities about Stepan Bandera. Bandera is admired by some Ukrainian nationalists for leading an independence drive before and during World War II. He is reviled by others, including many in Poland, because his movement sided with the Nazis and killed Poles and Jews.

The Peruvian outlet Extra reported that one such event did take place at a school. A Polish news report on the event then prompted Grzegorz Braun, a far-right Polish member of the European Parliament, to propose that people who profess support for Bandera be barred from entering Poland.

"The Russian fake thus both fed on, and fed into, historical tensions between Ukrainians and Poles," OpenAI wrote.

Darren Linvill, a co-director of Clemson University's Media Forensics Hub who studies online influence operations, said Russia is "constantly engaging in these sorts of active measures," including narrative laundering and fake personas built to push propaganda that matches Moscow's worldview.

Who was behind it

OpenAI did not attribute the campaign to a specific threat actor. It said the evidence suggests links to a Russian organization known as Politology or La Compania, which has been reported to be a successor to influence operations tied to the late Yevgeny Prigozhin and his Wagner Group.

The operators mostly prompted ChatGPT in Russian and appeared to be based in Russia. OpenAI does not allow access to its models from Russia, so they connected through VPNs.

The operators also leaned heavily on ChatGPT to draft reports to their own superiors. OpenAI said the Russian operation did this more than anything else, and that in both operations the actors used "questionable or outright deceitful methodologies" to exaggerate their effectiveness. The Peru claims come from those self-reported documents. OpenAI said it saw independent evidence of the school event through Extra and of the Polish reaction.

The Iranian side, briefly

The Iranian operation, labeled "Bogus Bylines," is the part already on the record. Seven personas claiming Western origins pitched long-form pieces to small and midsize outlets focused on geopolitics and the Middle East. Articles ran from July 2025 to October 2026, with frequency rising after the U.S.-Iran war began earlier this year.

OpenAI found that the campaigns focused on placing false narratives in established outlets appeared to reach a wider audience than the ones built on fake social media accounts. The Iranian operation's social media comments gained almost no traction.

Old playbook, new tools

OpenAI's own assessment is that both operations "closely resembled complex influence operations of the pre-AI age," with AI making some workflows easier. It pointed to "Alice Donovan," a fake journalist tied to Russian military intelligence whose work ran in Western outlets in 2016-17, and to the 2020 "PeaceData" outlet linked to people associated with the Internet Research Agency.

The takeaway from OpenAI's data is that the human-facing parts did the damage: real local staff, real news outlets, real politicians. The AI mostly handled drafting, translation and paperwork.

OpenAI has not identified any of the roughly dozen outlets that ran the Iranian bylines, and it has not said whether it notified them. Whether those publications pull the articles is still open.