OpenAI bans Russian propaganda unit that tricked schools in Peru with fake emails

OpenAI banned ChatGPT accounts tied to two covert influence operations — a Russian campaign it nicknamed "Dark Clark" and an Iranian one called "Bogus Bylines" — and rated the Russian operation at Category 5 on the six-point Breakout Scale, the first operation at that level it has disrupted since it began publishing these threat reports. According to OpenAI's new threat report, the Russian operators used ChatGPT to write internal reports and, in May 2026, set up a fake email address posing as Lima's Regional Directorate of Education to trick schools in Peru into holding events honoring Stepan Bandera on May 21, Peru's Day of Cultural and Linguistic Diversity; the Iranian operation reached Category 4. Both campaigns reached ChatGPT through VPNs, and OpenAI says it has shared information on both with the relevant authorities.

The OpenAI logo. Image: OpenAI logo, public domain via Wikimedia Commons OpenAI has banned ChatGPT accounts run by two covert influence operations, one from Russia and one from Iran, and says in a new threat report https://openai.com/index/disrupting-ai-enabled-false-front-operations/ that the Russian one ran a fake “think tank” in Latin America staffed by people who didn’t know who they were working for, and tricked schools in Peru into holding events honouring the Ukrainian nationalist Stepan Bandera. OpenAI calls both “false front” operations, planting messages through real institutions and editors rather than fake accounts. It rates the Russian campaign at Category 5 on the six-point Breakout Scale, which measures how far an influence campaign spreads, the first operation at that level it has disrupted since it began these reports. The Iranian one reached Category 4. Both reached ChatGPT through VPNs, and OpenAI, which last month started giving Ukraine its Daybreak cyber-defence system for free https://madrobot.blog/2026/09/23/openai-daybreak-cyber-defence-ukraine-free/ , says it has shared information on both with the relevant authorities. Fake emails to schools in Lima OpenAI nicknamed the Russian campaign “Dark Clark”, after “Mia Clark”, a made-up persona it says the operators used to run a self-styled research platform called the Social Research Center. They mainly used ChatGPT to write internal reports to an unnamed superior about campaigns against Ukraine and meddling in Argentina and Bolivia. In one, the operators said that in May 2026 they set up a fake email address posing as the Regional Directorate of Education in Lima and told schools to hold Ukraine events on Peru’s Day of Cultural and Linguistic Diversity, May 21, with explicit instructions to reference Bandera. Some Ukrainians see him as an independence figure; in Russia and Poland he is associated with fascism and wartime atrocities. According to the operators, some schools wrote back confirming the events, with pictures. The operators then claimed to have planted stories in Peruvian and Polish media accusing Ukraine of “exporting” ultra-nationalism. OpenAI found matching articles in Peru, Poland and an English-language outlet in Hungary, some quoting a Polish MEP who proposed that people showing “anti-Polishness” be declared persona non grata. In June, the operators said a second fake email got schools in Ecuador to hold a ceremony pledging allegiance to President Daniel Noboa and to Erik Prince, the former head of Blackwater. OpenAI found Ecuadorian coverage matching that claim, and a detailed rebuttal from Ecuador’s education minister. Fake audio, leaked letters and a polished contract The reports also describe faked audio: a clip attributed to Ukraine’s consul in Ecuador insulting Ecuadorians, and, at the height of anti-government protests in Bolivia in late May, a recording of a supposed worker at La Paz’s state water company, EPSAS, warning that the water would be shut off. EPSAS issued an official denial. In March, one operator asked ChatGPT to proofread a fake Spanish-language contract suggesting a front company was recruiting Ecuadorians to fight in Ukraine. A picture of the “contract” later spread online and was fact-checked in early April. OpenAI ties the operation to “Politology”, also known as “La Compania”, which open-source researchers describe as a successor to the Wagner Group and other ventures founded by Yevgeniy Prigozhin. The operators took credit for two 2024 campaigns against Argentina’s president Javier Milei, a false story that he bought Cartier jewelled collars for his dogs and paid-for anti-Milei graffiti in Buenos Aires, both already publicly linked to Politology through leaked documents. OpenAI warns the operators’ boasts can’t be taken at face value: they also asked ChatGPT to find events they could take credit for. Iran’s seven fake journalists The Iranian operation, which OpenAI calls “Bogus Bylines”, used ChatGPT to polish long English-language articles about the US-Iran war, check them against individual outlets’ submission rules and write pitch emails to editors under seven invented names: Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson. Most posed as Western freelance journalists. OpenAI found almost 100 articles under those bylines across more than a dozen small and medium outlets between July 2025 and October 2026, rising sharply after the war began; one outlet had almost 2 million Facebook followers. The operators also generated batches of replies on social media casting the US as the aggressor and Israel as an unreliable ally, but these drew little engagement. Their internal reports inflated their reach by counting views on the posts they replied to rather than on their own replies. OpenAI says the activity looked like a commercial, for-hire campaign but it couldn’t identify who ran it. Why it matters OpenAI’s finding is that these looked like pre-AI influence operations, with ChatGPT speeding up the reports, translation and editing rather than inventing new tactics. The campaigns that went furthest were the ones that got real schools, staff and editors to carry their material, which is a weakness no account ban fixes on its own. Sources: OpenAI, Disrupting AI-enabled “false front” operations https://openai.com/index/disrupting-ai-enabled-false-front-operations/