OpenAI Says Iranian ChatGPT Campaign Planted 100 Fake Articles in 20 U.S. Outlets
Updated
Updated · The Washington Post · Oct 9
OpenAI Says Iranian ChatGPT Campaign Planted 100 Fake Articles in 20 U.S. Outlets
3 articles · Updated · The Washington Post · Oct 9
Summary
About 100 AI-generated articles tied to users in Iran ran in at least 20 news outlets, OpenAI said, describing a for-hire influence campaign focused on U.S. military action in Iran.
Seven fake bylines, including Ervin Hoskins, were used to place English-language pieces that were refined with Persian prompts to match outlet guidelines and generate pitch emails to editors.
Port St. Joe Star publisher David Adlerstein said he ran one Hoskins column after it was forwarded by PeaceVoice, whose director said the group tightened its no-AI policy after OpenAI contacted him.
Daily Kos and Middle East Monitor removed articles linked to the fake personas, while OpenAI said it banned the accounts, shared information with authorities and found some profile photos were AI-generated.
The case underscores how AI tools can slip fabricated identities into real media channels, where even small outlets can give influence operations genuine readers and local reactions.
What hidden digital breadcrumbs allowed OpenAI to uncover these complex, VPN-cloaked foreign influence operations before they could cause further geopolitical damage?
How did a fake AI-managed persona successfully trick real researchers into producing legitimate content for a covert Russian influence campaign?
If AI can generate endless propaganda, why do sophisticated foreign networks still struggle to build genuine audience trust and distribute their deceptive content?
OpenAI’s Breakout Scale: How Generative AI Supercharged Russian and Iranian Influence Operations—and What It Means for Global Information Security
Overview
In October 2026, OpenAI disrupted two major state-linked influence operations: Russia’s 'Dark Clark' and Iran’s 'Storm-2035.' Russian operators used ChatGPT to create fake reports and documents, successfully deceiving Latin American schools and politicians, which led to OpenAI rating the campaign as a high-impact Category 5. Similarly, Iranian operators used ChatGPT to refine articles and social media content, placing over 100 pieces in real publications and earning a Category 4 rating. These AI-driven campaigns not only manipulated public opinion but also inflamed diplomatic tensions, such as between Ukraine and Poland, showing how generative AI is making global disinformation more effective and harder to detect.