IUVM: Iran-linked influence content network

OpenAI

OpenAI banned accounts associated with the Iran-origin operation "IUVM", using AI to generate and translate pro-Iran, anti-Israel and anti-US website content.

This case study was originally published in OpenAI’s May 2024⁠(opens in a new window) report.

Actor

We banned a small number of accounts linked to people associated with the International Union of Virtual Media (IUVM)⁠(opens in a new window), an Iranian entity that has been studied by the open-source community since 2018.

Behavior

This campaign targeted global audiences and focused on content generation in English and French. It used our models to generate and proofread article⁠(opens in a new window)s, headlines and website tags. This content was then published on IUVM’s current website, iuvmpress.co (earlier IUVM domains were seized by the FBI in 2020⁠(opens in a new window)).

The articles were typically created the day before they were published; the website tags were created immediately before publication and were likely automated. On one occasion, we identified a set of website tags that included our model’s response message, suggestive of automation (or poor proofreading).

Content

The content that this network generated consisted of long-form articles, headlines and website tags. This content was typically anti-US and anti-Israel and praised the Palestinians, Iran, and the “Axis of Resistance”.

Impact assessment

IUVM’s online presence has been reduced by repeated social media takedowns⁠(opens in a new window) and the FBI’s seizure of its domains⁠(opens in a new window). Beyond its website, as of May 23, 2024, we identified IUVM-branded accounts on TikTok, VKontakte and Odnoklassniki. These social media accounts had, respectively, 10, 76 and 274 followers or subscribers.