Iranian Operators Used ChatGPT to Plant ~100 AI-Generated Articles Across U.S. Publications
An Iranian influence operation used ChatGPT to produce almost 100 opinion and news articles critical of the U.S. war against Iran and placed them across roughly a dozen news outlets worldwide, OpenAI disclosed on October 8, 2026. Reporting by The Washington Post followed on October 9, putting the outlet count at at least 20. The company said it has identified and banned the ChatGPT accounts involved.
The operation used a fake byline — “Ervin Hoskins” is one confirmed pseudonym — to submit AI-generated content to outlets ranging from small local papers to larger digital publications. One article appeared in the Port St. Joe Star, a weekly Florida newspaper, warning that “the bill will follow Americans beyond the fighting itself.” The same byline has appeared in Daily Kos and Middle East Monitor, which together count more than 3 million Facebook followers.
Scale and Reach
OpenAI’s report, titled “Disrupting AI-enabled ‘false front’ operations,” counted almost 100 articles published or syndicated under deceptive bylines between July 2025 and October 2026. The company initially characterized placement as across “roughly a dozen” to “over a dozen” outlets worldwide. The Washington Post’s review put the figure at at least 20 news outlets; Hindustan Times reported “about a dozen publications” — counts that reflect both the scale of the campaign and which placements individual outlets later removed.
The outlets were international. Middle East Monitor, a UK-based publication, was among those that published the content. The articles focused on the economic and human cost of the ongoing U.S.-Iran conflict, framing that allowed the content to pass initial editorial review at outlets with limited fact-checking infrastructure.
OpenAI’s Response
OpenAI banned the accounts responsible and published its findings in a threat intelligence report titled “Disrupting AI-enabled ‘false front’ operations.” The operation adds to a documented pattern. OpenAI has previously identified and disrupted influence operations from China, Russia, and Iran using its models for content generation, translation, and distribution research.
The Structural Problem
What distinguishes this operation is placement. Generating AI content is the easy part; getting it into real, named publications with established audiences is where influence operations historically struggle. The Iranian campaign cleared that bar by targeting smaller outlets with looser editorial controls, then using the same bylines at larger outlets to build apparent credibility.
Jessica Brandt, who ran the Office of the Director of National Intelligence’s Foreign Malign Influence Centre until 2025, told CNN: “It’s remarkable that the Iranians managed to land their content in real media outlets, expanding the potential reach of their campaign.”
AI-assisted content generation does not require a novel jailbreak or API abuse. Standard ChatGPT access, at consumer pricing, is sufficient to produce volume. The bottleneck for influence operations has shifted from content generation to editorial infiltration — a problem that model safety and usage policies do not directly address.