OpenAI Cuts Off ChatGPT Accounts in Russian "Dark Clark" Operation Targeting Latin America
OpenAI has banned ChatGPT accounts linked to Russian and Iranian influence campaigns built on "false front" identities: a purported Latin American think tank with unwitting staff, and seven fictional Western journalists who had nearly 100 articles published or syndicated in roughly a dozen small and mid-sized online outlets.
OpenAI published its threat report and an accompanying blog post on Thursday, October 8, 2026, and the findings were covered the same day by NBC News, NPR, CNN, The Hill and AFP/Yahoo News. The company describes the Russian campaign — codenamed "Dark Clark" — as the most far-reaching influence operation it has uncovered and disrupted so far, according to NBC News. The Iranian operation has been codenamed "Bogus Bylines."
What's new
The concrete novelty in this disclosure is not that Russian and Iranian actors run influence campaigns — that is well established. What's new is two things: how far the Russian operation is claimed to have reached, and how elaborate the front identity was.
According to OpenAI, the operators behind "Dark Clark" ran a so-called think tank called the Social Research Center in Latin America as a front organization. Through it, they created forged leaked documents and scripts for audio recordings, with the aim of damaging Ukraine's reputation in the region and influencing local political outcomes, the company says.
"Available evidence indicates that SRC's employees in Latin America were not aware that they were working for a Russian group," OpenAI's report states, as quoted by AFP/Yahoo News.
The Peru gambit and the MEP reaction
The most striking individual thread in the Russian campaign runs via Peru to the European Parliament. According to OpenAI, the operators posed as a regional education authority in Peru and thereby got some schools to hold activities about Ukraine focused on Stepan Bandera — the Ukrainian nationalist who is a contested and highly charged figure in Polish historical understanding. NPR covered the matter, citing OpenAI.
News coverage of these activities in Poland is said, according to the operators' own reports, to have prompted a Polish member of the European Parliament to propose banning Bandera supporters from entry. The Hill writes, citing OpenAI, that the campaign is also claimed to have contributed to tensions between Ukraine and Poland.
But this is also where the sourcing becomes delicate: the most dramatic claims of impact — the news coverage, the MEP reaction — originate from the operators' own internal reports, written with ChatGPT. NPR points out that it is according to these reports that the operators claimed to have inserted false narratives into news media and provoked the reaction from the MEP. OpenAI itself notes that the actors used dubious or outright deceptive methods to exaggerate their own effectiveness. The impacts should therefore be read as the operators' own claims, not independently verified facts.
Bogus Bylines: seven fictional journalists
The Iranian operation was structured differently. Instead of one institutional front, the actors created seven journalist personas from Western countries. Under these pseudonyms, OpenAI identified nearly 100 articles that were published or syndicated in roughly a dozen online outlets around the world, in the period from July 2025 to October 2026, the company says, according to AFP/Yahoo News.
The outlets were small and mid-sized, not established mainstream media. But some have significant followings: CNN highlights Middle East Monitor, based in London, which has nearly two million followers on Facebook. In other words, the articles did not reach large, established outlets — but they did reach audiences through specialized and regional platforms with real readers.
What the AI was actually used for
The picture of ChatGPT's role is nuanced compared with earlier fears that AI would automatically generate mass content. In these campaigns, the tool was used as a workhorse for the operators: OpenAI says the accounts were used to draft internal reports, articles, pitches to newsrooms, and social media comments. In the Iranian campaign, some accounts were used with prompts in Farsi, according to CNN. The Hill notes that both operations made heavy use of AI to write internal reports — and that the Russian one was the most intensive on this point.
Many of these internal reports were thus the operators' own assessment of how well their work had gone — written with the same chatbot. That makes them questionable evidence, something OpenAI itself acknowledges.
Who was behind it — and what we don't know
Attribution is unresolved on the Iranian side. OpenAI writes in the blog post, as quoted by CNN: "The overall activity appeared consistent with a commercial actor running a for-hire influence campaign, but we are not able to identify the specific actor involved." That means the company can neither point to a concrete actor nor confirm any connection to Iranian authorities — even though the campaign has been described as Iranian, likely based on the actors' own markers.
On the Russian side, the targeting appears clearer: the campaign was aimed at Ukraine's reputation and at the relationship between Ukraine and Poland, with operations in Latin America. But even here it is important to distinguish what OpenAI can document (the accounts, the use of ChatGPT, the think tank front, the forged documents, the articles actually published) and what rests on the operators' self-reporting (the impacts, the MEP reaction, the reach).
Why it matters
The case illustrates how modern influence campaigns actually work in practice: not as autonomous botnets spreading content on their own, but as human-driven operations where AI shortens the distance between idea and finished product — forged leaks, credible-sounding journalist bylines, pitch emails to newsrooms, and internal reports documenting "results."
For OpenAI, the report is also part of the company's ongoing practice of publishing so-called disruption reports: the accounts are banned, the findings are made public, and the impacts are documented. This time, the company describes itself as having disrupted its most far-reaching operation so far. Open questions remain — in particular who was behind "Bogus Bylines," and how much of the reported impact actually took place.