Blog
OpenAI bans Russian and Iranian influence operations that used fake think tanks and journalists
On 8 October 2026 OpenAI said it banned ChatGPT accounts tied to a Russian operation running a fake Latin American think tank and an Iranian operation using seven fake journalist personas. It rated the Russian campaign Category 5, the first it has disrupted at that level.
Two operations, one pattern
On 8 October 2026, OpenAI published a threat report saying it had banned ChatGPT accounts tied to two covert influence operations, one from Russia and one from Iran. Both used AI alongside older tactics to run "false front" entities: a fake think tank in one case and fake journalists in the other.
OpenAI rated the Russian operation Category 5 on the IO Breakout Scale, which runs from 1 (lowest) to 6. It says this is the first Category 5 operation it has disrupted since it began reporting. The Iranian operation reached Category 4. OpenAI says it has shared information on both cases with the relevant authorities.
Russia: "Dark Clark"
OpenAI says the Russia-origin operators, connecting through VPNs, mainly targeted Latin America. Their aims were to damage Ukraine's reputation in the region and to interfere in local politics, especially in Argentina and Bolivia.
The centre of the operation was the Social Research Center (SRC), a self-described research platform run through a fake persona called "Mia Clark." According to OpenAI:
- The operators' own reports discussed pay scales, hiring and firing at the SRC, suggesting they controlled it.
- The SRC's staff in Latin America appear to have been unaware they were working for a Russian group.
- OpenAI found more than 60 articles on the SRC website, most of them original.
- OpenAI calls it "the most complex attempt to run a front identity" it has disrupted in two and a half years.
Most of the ChatGPT use was writing internal reports to superiors, not creating content. The reports described fakes the group claimed to have spread. OpenAI says it corroborated some of them through open sources. They include:
- A fake email to schools in Lima, Peru, asking them to hold events referencing Ukrainian nationalist Stepan Bandera. Matching stories appeared in Peruvian and Polish media, and OpenAI says a Polish Member of the European Parliament reacted.
- A fake audio clip of a Bolivian state water company worker claiming water to La Paz would be shut off. The company issued an official denial.
- A fake "contract" used to claim Ecuador's government was helping recruit Ecuadorians for Ukraine's army, which fact-checkers debunked.
OpenAI warns that the operators also took credit for events they had nothing to do with, so their claims of impact "cannot be taken at face value." The report does not attribute the operation to a specific Russian state body. It notes the operators asked about "Politology," a reported Wagner Group successor, far more than about any other Russian network.
Iran: "Bogus Bylines"
The Iran-origin operators used seven fake "journalist" personas, including "Ervin B. Hoskins" and "Michael Harrison." They used ChatGPT to polish long-form articles about the US-Iran conflict and draft pitch emails to editors.
OpenAI says it found almost 100 articles under these bylines across roughly a dozen small and medium online outlets, published between July 2025 and October 2026. One outlet had almost 2 million Facebook followers. The group also generated batches of social media comments in English and Persian, but OpenAI found little engagement on them.
OpenAI says the activity looked like "a commercial actor running a for-hire influence campaign" but could not identify who.
The takeaway
OpenAI's main point is that the AI part was not new. These looked like pre-AI influence operations, with AI making some tasks easier. Across the 30 covert operations it has exposed since early 2024, OpenAI says the ones that place content in real media outlets reach further than those relying on fake social accounts.
For editors and publishers, including small sites that accept guest posts, the lesson is practical: verify who a contributor really is before publishing them.