OpenAI bans two covert influence operations

On 8 October 2026 OpenAI reported that it had banned two covert influence operations that used its models to support "false front" entities: a Russian one it calls "Dark Clark", which ran a "think tank" in Latin America through unwitting local people, and an Iranian one, "Bogus Bylines", built on seven invented journalists. It rates the Russian operation Category 5 on the IO Breakout Scale, the first of that level it has disrupted, and the Iranian Category 4.

What the page says. OpenAI banned two operations that combined its models with more traditional techniques to support "false front" entities, through which they laundered geopolitical, conflict-related messaging into target audiences while hiding its origin. The Iranian one had seven "journalist" personas, which it used to pitch long-form articles to small and medium online outlets around the world, and generated batches of social media comments, mostly on the US-Iran war. The Russian one appears to have co-opted unwitting people in Latin America to run a "think tank" on the ground, created fake "leaked" documents and audio scripts, some of which OpenAI saw spreading online; both made heavy use of AI to draft internal reports and in both cases used questionable or outright deceitful methods to exaggerate the operators' effectiveness. The Russian operation, "Dark Clark". A cluster of ChatGPT accounts originating in Russia (OpenAI does not allow access from Russia, so they used VPNs) that targeted countries across Latin America. Much of the activity, the page says, appeared aimed at undermining Ukraine's reputation in the region, some at influencing local politics, especially in Argentina and Bolivia. The operators mostly asked ChatGPT to draft internal reports on their activity, to create content and to prepare reports on a self-described "research platform", the Social Research Center, which they appear to have controlled through a fake persona, "Mia Clark". Some of the fakes, OpenAI says, have been attributed by open-source researchers to a Russian entity known as "Politology" or "La Compania", reported to be a successor to the Wagner Group, and the operators asked the models to translate or explain public reporting about Politology and Wagner far more than about any other Russia-origin network. OpenAI calls the operation unusual in two respects: it appears to have co-opted people in Latin America to work for the Social Research Center, who did not know they were working for a Russian group, and its staff produced most of the original content; and open-source evidence suggests its fakes spread widely enough to provoke fact checks and official denials, one fake planted in Peru even feeding into public tensions between Ukraine and Poland. OpenAI shared the information with the relevant authorities. Assessment. On the IO Breakout Scale (1 to 6) OpenAI places the Russian operation in Category 5, the first it has disrupted since it began reporting, and the Iranian one in Category 4; both landed content in mainstream media outlets, not only on social media. Among the 30 covert influence operations it has exposed since early 2024, those that try to land content in real media outlets tend to have the highest potential reach. What the record does not claim. All of this is OpenAI's assessment of its own investigations; no independent confirmation was read. Of the page's roughly 34,000 characters the introduction and the section on the Russian operation (about 9,500) were read closely, the section on the Iranian operation only through the introduction. Of the Russian operation the page says that in most cases OpenAI did not see the models used to create campaign content, only to report on it; it does not say this of the Iranian operation, where the models wrote articles and comments.