On October 8, 2026, OpenAI announced that it had banned a cluster of ChatGPT accounts originating in Iran and linked them to an influence operation it named Bogus Bylines. The company said operators used seven fictitious journalist identities to pitch articles to online publications and identified almost 100 articles published or syndicated under those bylines across roughly a dozen outlets.
NeoTeo previously covered a separate Russia-origin ChatGPT operation; Bogus Bylines is the Iran-origin case.
Seven fictitious bylines and almost 100 articles
OpenAI identified seven bylines: Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson. The personas generally presented themselves as Western journalists, usually Americans; Noah Lamington’s biography claimed a New Zealand location.
The articles focused primarily on the geopolitics of the US-Iran conflict. OpenAI placed the identified publications between July 2025 and October 2026. It said their frequency increased after the conflict began in 2026.
How ChatGPT assisted the operation
OpenAI said operators used ChatGPT to review article drafts against individual outlets’ submission requirements, suggest edits and write emails pitching the articles to editors. The workflow extended beyond article placement: the company also described ChatGPT-generated batches of social-media comments in English and Persian, as well as assistance with internal reports.
Some comments praised or endorsed articles linked to the operation; others addressed the conflict and US politics. Replies OpenAI located generally drew little visible engagement, often in the single or double digits, and made up a minority of comments on the posts examined.
What OpenAI’s two ratings describe
OpenAI rated the article-placement activity Category 4 and the social-commenting activity Category 2 on its six-category IO Breakout Scale. The ratings are the company’s assessments of those two strands of activity, not independent measurements of persuasion or audience impact.
The higher article-placement rating concerned almost 100 articles published or syndicated across roughly a dozen outlets. The lower social-commenting rating applied to batches of replies that OpenAI said generally received little visible engagement.
OpenAI’s assessment of the operator
OpenAI said the activity appeared consistent with a commercial actor running a for-hire influence campaign, but the company could not identify the actor.
