OpenAI said on Thursday that it disrupted an Iranian-linked influence operation that used ChatGPT to place almost 100 AI-generated articles under fake Western journalist bylines in real publications around the world, including several United States outlets.
The campaign, which OpenAI dubs a "false front" operation, is detailed in a report the company published on October 8 and in follow-up reporting by The Washington Post, CNN, NPR and Radio Free Europe/Radio Liberty. It is one of the clearest documented cases to date of a foreign influence network using a mainstream commercial chatbot not just to draft propaganda, but to manage the full editorial workflow of planting it in legitimate media. For more context on this story, see our ongoing AI industry coverage.
A Journalist Who Never Existed
At the center of the operation was a persona named Ervin B. Hoskins, presented as an American freelance journalist. According to RFE/RL, articles attributed to Hoskins cast the United States as the aggressor in the ongoing conflict with Iran, portrayed Israel as an unreliable ally, and framed Iran as a defiant underdog. Hoskins, OpenAI confirmed, never existed.
He was one of seven fabricated journalist personas, all purporting to be from Western countries, that operators used to pitch and place articles. "We identified almost 100 articles published or syndicated under the operation's bylines across roughly a dozen online outlets around the world," OpenAI said in its report, describing the targets as small to medium outlets generally focused on international affairs, geopolitics and the Middle East.
The Washington Post, which first reported the campaign's penetration of US media, found that a small Florida weekly published one of the fake op-eds. Other flagged placements appeared in outlets including Middle East Monitor, which told CNN it had removed the articles in question and was reviewing its submission process.
How ChatGPT Was Used
According to OpenAI's disclosure, the operators used ChatGPT in both Farsi and English across the entire placement pipeline: editing long-form article drafts, tailoring pitches to individual publications' editorial requirements, preparing outreach emails to editors, and generating batches of social media comments designed to look like independent reader reactions.
One example traced by researchers: an op-ed titled "Jared Kushner and the Privatization of America's Iran Strategy," published on June 12 by PeaceVoice, an op-ed distribution service run by the Oregon Peace Institute. In the following days the same article was republished by CounterPunch and LA Progressive, according to RFE/RL.
OpenAI said publication activity increased after the conflict between the United States and Iran began earlier this year. The company said it has banned the ChatGPT accounts involved in the campaign and rated the operation's ability to place content in outside editorial outlets as Category 4 on its six-point influence operation scale, according to CNN reporting.
Attribution Remains Open
Notably, OpenAI said it could not identify the specific actor behind the campaign or confirm direction from the Iranian government. Its assessment, as reported by CNN, pointed toward a commercial, for-hire influence operation rather than a directly state-run one — a structure that makes attribution harder and disruption less conclusive. Banning the accounts disrupted what OpenAI could observe; it did not establish that the broader network was dismantled.
Jessica Brandt, who led the Office of the Director of National Intelligence's Foreign Malign Influence Center from 2023 to 2025, told CNN the operation showed how AI helps foreign actors conceal their involvement and route messaging through authentic channels.
The Iranian campaign was not the only activity OpenAI flagged. The company's report also covered disruption of Russian-linked influence operations that used ChatGPT, as The Record from Recorded Future News reported. Separately, RFE/RL noted that Anthropic has previously disrupted Iranian efforts to misuse its Claude models for propaganda and surveillance of dissidents.
Publication Is No Longer Proof of Authorship
For newsrooms, the case lands on an uncomfortable truth: a byline, a coherent beat, a plausible biography and a polished pitch no longer imply that a contributor exists. Editors who rely on freelance submissions — particularly op-ed and commentary desks, which by design publish outside perspectives — now face a verification problem that spam filters and writing-quality heuristics cannot solve.
Practical responses flagged in coverage of the report include stronger contributor authentication, documented pitch histories, scrutiny of bylines that appear across multiple outlets in short windows, and provenance tooling that ties submissions to verifiable identities.
The economics also matter. An operation that once required a staff of writers can now be run by a small team directing a chatbot, with the model handling everything from translation to tone-matching a target publication's house style. OpenAI's report suggests the barrier to running a multi-persona, multi-outlet placement campaign has dropped to the point where a for-hire shop can do it as a commercial service.
None of the outlets named in the reporting were major mainstream news organizations, but several carried real audiences. That is the point, researchers say: placement in any credible publication moves a message into wider circulation through syndication, aggregation and search — which is exactly what the campaign achieved before it was disrupted.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
