OpenAI reports disrupting a covert Iranian influence operation

OpenAI said it had identified and banned a cluster of ChatGPT accounts tied to a covert Iranian influence operation, publishing the takedown as part of its recurring disclosures of state-linked misuse of its models. Microsoft’s threat-intelligence team had flagged related activity, tracked as Storm-2035, earlier the same month.

The accounts used ChatGPT to draft both long-form articles and short social media comments on divisive political topics — the war in Gaza, Israel’s presence at the Paris Olympics, and the 2024 US presidential election — in English and Spanish, then distributed the material through social accounts and websites styled to resemble genuine news outlets. Operators interspersed the political content with unrelated posts about fashion and beauty, apparently to make the accounts look more like ordinary users. OpenAI reported that the operation had achieved little reach: the identified posts drew few or no likes, shares or comments.

The disclosure fit a pattern OpenAI and other labs had established through 2024 of publishing periodic reports on state-linked misuse — covering Russian, Chinese and Iranian operations among others — as a form of public accountability distinct from, and generally more transparent than, the platforms on which the content was actually distributed. The muted engagement OpenAI reported was consistent with independent assessments through the year that AI-generated influence content was not yet demonstrably more persuasive or effective than operations built without it, even as the volume of such campaigns increased.