The latest disclosure came Thursday, when OpenAI said it disrupted an Iran-origin influence operation that used ChatGPT to help fake journalist personas place almost 100 articles in roughly a dozen online outlets around the world.
OpenAI dubbed the campaign “Bogus Bylines.”
The operators used seven fake journalist identities to pitch English-language articles, primarily about the US-Iran conflict. ChatGPT was used to refine drafts according to outlets’ submission requirements and prepare emails pitching them to editors.
The personas presented themselves as Western journalists, usually Americans. One, Ervin B. Hoskins, described himself as an “American freelance writer” and maintained accounts on X, Facebook, Instagram and Threads.
But OpenAI said platform transparency information showed the accounts were operated from Iran.
OpenAI identified articles under the operation’s bylines dating from July 2025 through October 2026 and said publication increased significantly after the US-Iran conflict began.
Some outlets subsequently promoted the articles through their own social media accounts, potentially exposing the material to audiences the operators could not otherwise reach.
The operators also used ChatGPT to produce batches of social media comments. Some were generated in response to posts promoting their own articles and praised or endorsed the stories, which OpenAI assessed was likely intended to make them appear more popular.
Other prompts requested comments portraying the United States as the aggressor in the conflict, Israel as an unreliable US ally and Iran as a defiant but resilient victim.
The operation generated more than two dozen batches of hostile Persian-language responses to posts by Iran International.
OpenAI said the replies it found generally received low engagement and rated that component Category 2 on its six-level IO Breakout Scale. The article-placement campaign was more successful, reaching Category 4 after repeatedly getting material published by media outlets.
OpenAI said the overall activity was consistent with a commercial actor running an influence campaign for hire but said it could not identify the actor behind it.
From influence to military targeting
The OpenAI findings follow a September report from Anthropic documenting Iranian-linked use of Claude for activities extending beyond influence operations.
Anthropic said it disrupted an “Iran-nexus threat actor” that used Claude to collect and analyze publicly available information to develop targeting recommendations against US naval forces in the region.
The actor used Claude to help construct a Python pipeline for tracking naval positions and compiling targeting handbooks.
The material included US personnel identified from captions on publicly available military photographs, ship and aircraft transponder identifiers, scripts for querying commercial satellite imagery and lists of public websites exposing US naval movements.
The actor also directed Claude to research vulnerabilities in shipboard systems, including maritime satellite communications terminals, Cisco communications equipment and industrial control products.
OpenAI and Anthropic have documented misuse by actors linked to several countries, including for cyber operations, influence campaigns and intelligence gathering, suggesting generative AI is increasingly being incorporated into activities that long predate the technology.
OpenAI concluded that the influence operations it uncovered resembled campaigns from the pre-AI era, with artificial intelligence making parts of the workflow easier.
The two investigations suggest AI is being used less to invent entirely new forms of influence and intelligence activity than to accelerate existing ones: drafting and pitching articles, manufacturing online responses, processing surveillance data, writing software and assembling open-source military intelligence.