OpenAI Disrupts AI-Enabled 'False Front' Influence Operations in Russia and Iran
Key point
OpenAI disrupted two sophisticated influence operations using AI to create fake personas and launder geopolitical narratives, marking the first Category 5 breakout in its reporting history.
Details
OpenAI announced the disruption of two AI-enabled influence operations (IOs) originating from Russia and Iran, part of a broader effort that has exposed 30 covert IOs over the past 2.5 years. These operations combined AI with traditional techniques to launder geopolitical messages through sophisticated false front organizations. The Russian operation, dubbed Operation Dark Clark, was rated Category 5 on the IO Breakout Scale—the first such rating since OpenAI began reporting—while the Iranian operation, Operation Bogus Bylines, was rated Category 4.
Operation Dark Clark: Russia
The Russian operation targeted Latin America with the goal of damaging Ukraine's reputation and influencing local political outcomes in countries like Argentina and Bolivia. It was controlled by a Social Research Center (SRC) in Latin America, which used a fake persona named Mia Clark. SRC employees were likely unaware they were working for a Russian group, serving as unwitting cutouts similar to the PeaceData case exposed by Meta in 2020.
Key activities included:
- Argentina (2024): Spreading misinformation about President Javier Milei, such as claims he bought a Cartier necklace for his dog.
- Peru (May 2026): Impersonating the Lima Education Department to instruct schools to hold events mentioning controversial Ukrainian nationalist Stepan Bandera, leading to actual events and media coverage.
- Ecuador (June 2026): Sending fake emails demanding loyalty pledges to President Daniel Noboa and Erik Prince, and spreading false audio claiming a water supply cutoff in La Paz, Bolivia.
AI was primarily used for internal reporting, content generation, and adapting language to match local accents and institutional styles, though native speakers were also involved.
Operation Bogus Bylines: Iran
The Iranian operation used ChatGPT to refine long-form articles and generate social media comments related to the US-Iran conflict. Operators created 7 fake journalist personas (e.g., Ervin B. Hoskins, Michael Harrison) posing as Western journalists. These personas pitched articles to over 12 online media outlets, with publication frequency spiking after the outbreak of the US-Iran conflict.
- Article Distribution: Approximately 100 articles were published or syndicated between July 2025 and October 2026.
- Social Media Comments: Operators generated large volumes of comments in English and Persian, often praising their own articles or promoting hostile narratives against Israel and the US. However, these comments received low engagement, with most likes coming from operator-linked accounts.
- Impact: While the social media commenting effort showed low engagement, the journalist persona activity achieved a Category 4 breakout, successfully placing content in outlets with significant followings (e.g., one outlet had nearly 2 million Facebook followers).
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.