OpenAI Reports Iranian AI Influence Operation Targeting US Media
OpenAI has reported that Iranian operatives used its artificial intelligence models and fake journalist identities to help place articles critical of the United States’ war against Iran in multiple media outlets. The disclosure, published on October 8, 2026, raises concerns about how AI tools can support covert influence campaigns by helping operators prepare content for publication under misleading identities.
According to OpenAI, the operation involved seven fabricated journalist personas used to pitch long-form articles to online publications. The company identified nearly 100 articles published or syndicated across roughly a dozen outlets worldwide. The content focused primarily on the US-Iran conflict, presenting narratives critical of Washington’s actions and seeking to reach audiences through established media channels.
The operation, which OpenAI named “Bogus Bylines,” involved ChatGPT accounts originating in Iran. The operators submitted prompts in Persian and generated material in both Persian and English, while using virtual private networks to obscure their location. OpenAI said the accounts were used to refine articles and editorial pitches, generate social media comments and prepare internal reports about the campaign.
Rather than relying exclusively on websites created to distribute propaganda, the operators sought publication through existing online outlets. This approach potentially allowed the material to reach readers through familiar media platforms, where articles may appear alongside independently produced reporting. However, the number of publications involved does not establish how many readers encountered the content or whether it changed public opinion.
How AI Helped the Fake Journalist Operation
OpenAI said the operation used its models to review and improve English-language articles about the geopolitical conflict, including checking drafts against specific publications’ submission requirements. The AI assistance helped refine editorial pitches and written material, reducing some of the work required to prepare articles for external publication. The disclosure does not mean the AI system independently selected the publications or orchestrated the entire operation.
The campaign also generated batches of social media comments, most of which concerned the US-Iran conflict. Some comments addressed US politics in the broader context of the war. OpenAI found little evidence that these comments attracted substantial engagement through likes, shares or replies, suggesting that this part of the operation had limited visible reach.
The articles had a broader distribution footprint than the social media comments. OpenAI identified nearly 100 articles across approximately 12 outlets, although the company said the publications were generally small or medium-sized platforms focused on international affairs, geopolitics and Middle Eastern issues. Publication or syndication alone, however, does not prove that every outlet knowingly participated in the influence campaign.
OpenAI said it banned the ChatGPT accounts associated with the operation and shared information about the case with relevant authorities. The company also said the activity appeared consistent with a commercial operation running an influence campaign for hire, but it could not identify the specific organization or individual behind it. The identity of the ultimate sponsor therefore remains unresolved in the company’s public account.
What the Disclosure Means for AI and Online Media
The case illustrates how generative AI can support several stages of a covert influence operation, from drafting and editing articles to preparing social media material. Such tools can make it easier to produce content in multiple languages and adapt it for different audiences. They do not, by themselves, guarantee publication, audience engagement or a successful attempt to influence public opinion.
The incident also highlights a challenge for news organizations: identifying deceptive authorship when submitted articles are polished, tailored to editorial guidelines and presented under plausible professional identities. Publications may need to strengthen contributor verification, examine suspicious submission patterns and establish clear procedures for checking authors’ backgrounds and affiliations.
At the same time, the findings require careful interpretation. OpenAI’s report documents the company’s investigation and actions against accounts using its services, but the available evidence does not establish that all articles were false in every factual detail. The central concern is the alleged use of fabricated identities and coordinated messaging to conceal the operation’s origins and purpose.
The disclosure is part of a wider pattern of AI-assisted influence activity that OpenAI has reported in previous years, including operations linked to Iran that generated articles and social media posts. The latest case demonstrates that disrupting accounts does not necessarily reveal every participant in a campaign or establish its full reach. Independent investigation remains important when assessing who organized such activity and how its content circulated.
For publishers, readers and technology companies, the key issue is not simply whether an article was written with AI assistance. It is whether the material uses deceptive identities, conceals sponsorship or misrepresents its origins to influence an audience. OpenAI’s findings show how AI can be incorporated into established influence tactics, while leaving important questions about the campaign’s organizers, actual readership and wider impact unanswered.
About The Author
