One ChatGPT account linked two Iranian influence ops the world thought were separate

OpenAI’s latest threat report details a small but telling crack in the wall between two Iranian influence operations the security community has long treated as distinct efforts. The company banned five ChatGPT accounts in February 2025 that were generating articles and social media posts for the International Union of Virtual Media (IUVM) and a cluster Microsoft tracks as STORM-2035. The real surprise isn’t the content—it’s the operator overlap.

One account was used to create material for both groups. It churned out French-language copy for critiquepolitique[.]com, a site publicly linked to STORM-2035, and English-language articles for iuvmpress[.]co, an IUVM property Reuters first exposed back in 2018. Public reporting had never connected those two domains before. That single account bridges a gap the open-source intelligence world missed, suggesting either a shared freelancer or a more centralized content farm than anyone assumed.

The operators weren’t particularly sophisticated, but they were careful. OpenAI found that the person running the dual-purpose account used its models to generate drafts, then manually rewrote the text before publishing. A semantic similarity analysis flagged the published versions as highly likely rewrites of the AI-generated originals—a cheap, low-effort evasion tactic that probably would have worked against simpler detection methods.

As for impact, this operation flopped. The al-Sarira X account, linked to STORM-2035, had 157 followers while following 895 accounts as of mid-January. Most tweets got single-digit engagement. On the Breakout Scale, OpenAI pegs this at the low end of Category 2: content splashed across multiple platforms, but zero evidence real humans cared. The narratives were the standard menu—pro-Palestinian, pro-Hamas, anti-Israel, anti-US—with a weird pivot to defending Bashar al-Assad during his regime’s collapse. Some banned accounts spent more time asking ChatGPT to design ESL lesson plans than to write propaganda, a detail that echoes earlier Iranian threat activity tied to individuals with English-teaching backgrounds. It’s a reminder that influence operations are often side gigs, not shadowy full-time jobs.