OpenAI said on October 8, 2026, that it banned clusters of ChatGPT accounts associated with Russia-origin and Iran-origin influence operations. The company described different workflows: one used ChatGPT most often for internal reporting, while the other used it to refine articles, pitch publications and generate social media comments. OpenAI’s report also assigned separate ratings to the Iranian operation’s article placements and comments.
A research front and fabricated bylines
OpenAI called the Russia-origin operation “Dark Clark” and said its operators appeared to run the Social Research Center (SRC), a Latin America-facing research front, through a fabricated persona named Mia Clark. The campaign mainly sought to undermine Ukraine’s reputation in Latin America; some activity appeared aimed at local political outcomes in Argentina and Bolivia.
OpenAI identified more than 60 articles on the SRC website, most of them apparently original. It said Latin American staff produced most of that original content and appeared to work in good faith, possibly without knowing about the Russian connection.
The Iran-origin operation, which OpenAI nicknamed “Bogus Bylines,” used seven fabricated journalist identities. OpenAI identified almost 100 articles published or syndicated under those bylines across roughly a dozen outlets. The operation also generated batches of social media comments.
How ChatGPT featured in each operation
The roles differed. OpenAI said internal reporting was the most frequent observed ChatGPT use in the Russia-origin operation; other uses included campaign material and reports about the SRC. In most cases, OpenAI did not observe its models being used to create campaign content.
For the Iran-origin operation, OpenAI described ChatGPT use to refine long-form articles, tailor submissions to publication requirements, draft pitches and generate comment batches. That distinction matters: the report describes AI assistance across different parts of the workflows, not a single method of producing every campaign item.
| Activity | ChatGPT use described by OpenAI | Distribution route | OpenAI rating |
| Russia-origin operation, “Dark Clark” | Internal reporting was the most frequent observed use; other uses included campaign material and SRC reports. | SRC website articles and social media activity. | Category 5 of 6 |
| Iran-origin article placement, “Bogus Bylines” | Article refinement and publication pitches. | Articles published or syndicated across roughly a dozen outlets. | Category 4 of 6 |
| Iran-origin social comments, “Bogus Bylines” | Generation of comment batches. | Comments posted on social media. | Category 2 of 6 |
What the ratings and engagement indicate
OpenAI’s six-point IO Breakout Scale ratings describe its assessment of the operations’ activity. The company rated “Dark Clark” Category 5, Iranian article placement Category 4 and Iranian comments Category 2. These ratings do not measure whether audiences were persuaded.
OpenAI said Iranian comment activity generally drew little engagement. It also described a weakness in operators’ own impact estimates: they counted views on original posts before and after posting replies, rather than measuring views generated by the replies. Those counts could not isolate the comments’ contribution.
OpenAI also said Russian operators sometimes sought credit for incidents they might not have caused. That caveat is particularly relevant when interpreting claims about campaign impact: a reported event and an operator’s claim of responsibility are not, by themselves, proof that the operation caused it.
