OpenAI removes Iran-linked accounts behind fake journalist campaign
Seven fabricated writers used ChatGPT to tailor articles and pitches that reached more than a dozen publications.
OpenAI said it had shut down ChatGPT accounts used by an Iran-based influence operation that created seven fake journalist identities and placed or syndicated nearly 100 articles through more than a dozen outlets around the world. The campaign, which OpenAI called “Bogus Bylines,” used fabricated bylines and social-media profiles to make political material appear to be independently authored.
The company disclosed the activity on Oct. 8. CNN said the accounts used prompts in Farsi to revise English-language articles and pitches, and to generate purported public reactions on social media.
The operators submitted drafts to ChatGPT, asked the model to assess them against particular publications’ submission rules, approved revisions and requested emails to editors. They also used it to write at least one professional biography, for the persona Michael Harrison.
The identities were Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson. Their biographies generally presented them as Western freelance writers or commentators interested in international affairs, human rights and conflict. Most purported to be American, while Lamington was presented as living in New Zealand.
The earliest article OpenAI identified appeared in July 2025. Publication accelerated in March, after the outbreak of war involving Iran, Israel and the United States, with articles focusing on that conflict, international politics and the consequences of US military action. Examples included “Is America’s Iran Policy Being Privatized? The Jared Kushner Question?” and “The Iran War and the Collapse of Western Moral Credibility.”
The campaign’s social profiles reinforced the false identities. Harrison appeared to have accounts on Medium, X and Facebook, while Hoskins had profiles on X, Facebook, Instagram and Threads. Transparency information on Hoskins’s accounts indicated activity from Iran despite his claimed identity as an American writer; Meta removed his Instagram account in August.
The operation placed material through ordinary publishing channels. A Hoskins column reached Florida’s Port St. Joe Star via PeaceVoice, a long-standing material supplier to the newspaper. Publisher David Adlerstein said he had not separately verified the purported author because of that established relationship. PeaceVoice said it had been unaware of the deception and subsequently tightened its rules for AI-generated submissions.
Daily Kos said the relevant items appeared on its community platform, where users can publish without the advance scrutiny applied to staff writers. Middle East Monitor said it had removed the flagged articles and was reviewing its submissions process. Some publications also promoted the articles through their own social accounts, potentially extending their reach beyond the network’s own profiles. Middle East Monitor alone had nearly 2 million Facebook followers.
After publication, the operators asked ChatGPT to create English- and Persian-language comments that would look like public responses. The requested posts portrayed the United States as an aggressor, criticized Israel and challenged US military policy. More than two dozen batches of hostile Persian-language replies targeted Iran International, while screenshots of posts promoting the operation’s own articles were used to request favourable responses.
OpenAI found little genuine engagement with identifiable comments, some of which were mainly approved by accounts associated with the operation. It rated the commenting component Category 2 on its six-level Breakout Scale, but rated article placement Category 4 because material had entered independent publications with potentially substantial audiences, rather than because it had shown an effect on public opinion.
OpenAI assessed the activity as consistent with a commercial influence operation run for hire, but could not identify the responsible organization. It also said the political content did not establish whether the Iranian government directed the campaign.