OpenAI published a report on 8 October 2026 describing two influence operations it had disrupted, one run from Russia and one from Iran. Both used its models to support false-front entities. OpenAI rated the Russian operation Category 5 on its scale of 1 to 6, the first Category 5 operation it has disrupted since it began reporting, and rated the Iranian operation Category 4.
The Iranian operation, which OpenAI calls "Bogus Bylines," used seven "journalist" personas to pitch long-form articles to small and medium online outlets around the world. It prompted ChatGPT in Persian and generated content in Persian and English. The operators hid their location with VPNs and used "American freelance writer" personas, but transparency settings still placed the accounts in Iran. OpenAI counts almost 100 articles across roughly a dozen outlets, and says both operations managed to land content in mainstream media rather than only posting it on social platforms.
The Russian operation, which OpenAI calls "Dark Clark," ran a "think tank" called the Social Research Center. OpenAI says it appears to have co-opted unwitting people in Latin America to run that think tank on the ground. The operation also made fake "leaked" documents and audio scripts. OpenAI groups its work into three streams, one of which was efforts to denigrate Ukraine and undermine recruitment for the Ukrainian Armed Forces. One fake the operation claimed to have planted in Peru fed into tensions between Ukraine and Poland. Open-source searches found matching stories in Peruvian and Polish press, and in an English-language publication in Hungary.
What the operators did with ChatGPT
The operators used ChatGPT for three main tasks, the first of which was writing internal reports on their activities. Those reports claimed fake stories had spread across Latin America. OpenAI says the claims were possibly made to inflate the operators' apparent effectiveness, and it describes both operations as using questionable and outright deceitful methods to exaggerate their results.
The Russian operators used a persona called "Mia Clark" as part of the front. OpenAI says the people it describes as co-opted were not aware they were working for a Russian operation. OpenAI says it has shared information on the case with the relevant authorities, and that the overall activity appeared consistent with commercial work.
What the independent coverage adds
The Hill reported the announcement separately from the AFP copy carried by RFI and Al-Monitor, so it is the one independent account in the sources checked. None of the three measures how far either operation reached, and none names the outlets involved. For editors, the more useful part of OpenAI's post is the method: personas that pitch long-form pieces, transparency settings that point back to Iran, and the use of VPNs. Newsrooms that take freelance or syndicated pitches can check bylines against that method. The report describes how the operations worked, not a full list of names, so it is a set of signs to check rather than a list of accounts to block.