What Did OpenAI Find?
On October 8, 2026, OpenAI published a report on two influence operations it had banned from ChatGPT, one from Russia and one from Iran, that ran what it calls “false front” entities (OpenAI). The Iran-origin operation used seven bylines to pitch long-form articles about the US-Iran conflict to online outlets, and OpenAI named it “Bogus Bylines.” The report lists the names: Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson. According to bios posted with some articles, each claimed to be a Western journalist. Some were backed by social media accounts; OpenAI says the Hoskins accounts showed locations in Iran.
OpenAI’s own page did not load for us directly, so we read an Internet Archive copy from October 9. The report is now nine days old. This guide explains what it shows for contributed articles rather than reporting it as new.
Category on OpenAI's IO Breakout Scale (1 lowest, 6 highest)
Ratings are OpenAI's own assessment. OpenAI calls Dark Clark the first Category 5 operation it has disrupted. It rated the two Bogus Bylines workstreams separately because, it says, their impact "varied significantly."
The chart shows OpenAI’s own impact ratings. It rated the article placements Category 4, “based on its consistent breakout to a number of media outlets,” and the social media comments Category 2, saying they drew few likes or replies. OpenAI wrote that AI “can give such false-front operations greater scale, efficiency, linguistic fluency, and editorial ability.”
How Did the Articles Get Published?
Through the same route a real contributor uses. OpenAI says the operators asked ChatGPT to assess English-language drafts “against the submission criteria for a specific online outlet,” suggest edits, and then write an email “to pitch the article to the editor of the online outlet in question.” For one persona, they had the model write the bio. OpenAI says the operation used its models to “refine” articles and does not say who wrote the first drafts.
Radio Free Europe/Radio Liberty traced one example. A June 12 article under the Hoskins byline ran on PeaceVoice, an op-ed distribution service, and over the next few days on CounterPunch and LA Progressive (RFE/RL). That is syndication working as designed: one accepted piece carried a byline to several more sites. Our guide on contributor networks covers how such programs vet writers.
Which Outlets Were Affected, and What Did They Say?
OpenAI does not name the outlets. It describes them as “small to medium outlets, generally focused on international affairs, geopolitics, and events in the Middle East, sometimes from an antiwar or progressive point of view,” and says one had almost 2 million Facebook followers. RFE/RL named PeaceVoice, CounterPunch and LA Progressive for the Hoskins article and said articles also appeared in libertarian-leaning publications. VOA, citing CNN, named Middle East Monitor, a London-based site with nearly two million Facebook followers. Its editor-in-chief, Dawoud Abdullah, told CNN he would review the articles and “take appropriate action”; another representative later said the articles had been taken down and the outlet was reviewing its submissions process (VOA News). We found no response from the other outlets in the sources we read.
Where Do the Accounts Disagree?
On three points.
- Outlet count. OpenAI’s summary says “roughly a dozen” outlets, and its detailed section says “over a dozen different publications.” VOA wrote 12. We use OpenAI’s wording.
- Dates. VOA wrote that the articles ran “from July 2025 to September.” OpenAI says the earliest article it found was published in July 2025 and the latest in October 2026, with far more after the US-Iran conflict began.
- Who was behind it. VOA described “Iranian regime agents,” and RFE/RL wrote of “Iranian operatives.” OpenAI says the accounts originated in Iran and the activity “appeared consistent with a commercial actor running a for-hire influence campaign, but we are unable to identify the particular actor involved.” It says it shared the information with the relevant authorities.
There is also a wording difference inside the report. Its overview says both operations landed content “in mainstream media outlets, rather than simply posting it on social media,” while its Iran section calls the outlets “small to medium.” VOA wrote that none of the outlets appeared to be mainstream news organizations.
Can an Editor Check Who a Contributor Is?
The report shows the usual signals are easy to fake: a plausible name, a bio, social accounts and a clean pitch. It does not say what checks the outlets ran. Our guide on whether editors will now verify who wrote your op-ed looks at what verification can involve, and author bylines covers what a byline signals. No source in this case sets a verification standard, so what follows is our recommendation, not a rule.
What This Means for Anyone Submitting or Buying a Contributed Article
A real contributor should expect more requests for proof and should offer it first: a video call, a profile with a work history, an institutional email address and a phone number. If you hire a vendor to place a contributed article, ask who the named author is, whether that person wrote or approved the piece, whether payment is disclosed as sponsored where it applies, and how the outlet vets contributors. Never publish under an invented byline or use a synthetic headshot for a real client; the outlet and its readers are the ones deceived, whatever the motive. See How to Write a Contributed Article for a Trade Publication and How Many Expert Pitches Are Written by AI?
Methodology and Limitations
Four limits apply. First, we read OpenAI’s report through an Internet Archive copy because the live page did not load for us. Second, the counts are OpenAI’s, and its own wording varies between “roughly a dozen” and “over a dozen” outlets. Third, outlet responses come secondhand through VOA’s account of CNN’s reporting. Fourth, we take no position on the politics of the articles and did not assess OpenAI’s attribution.
Conclusion
By OpenAI’s account, a set of invented bylines placed close to 100 articles by doing what ordinary contributors do, with an AI tool making each pitch fit the outlet. For anyone who writes or buys contributed coverage, the practical response is to make your own authorship easy to confirm and to favor outlets that check it.
Frequently Asked Questions
What is OpenAI's "Bogus Bylines" report?
It is part of OpenAI's October 8, 2026 report on "false front" influence operations. OpenAI says an Iran-origin operation used seven invented journalist bylines to place almost 100 articles in roughly a dozen small and medium online outlets.
How did ChatGPT help place the articles?
OpenAI says the operators asked ChatGPT to check English-language drafts against each outlet's submission criteria, suggest edits, and write the pitch emails to editors.
Did OpenAI say the Iranian government was behind it?
No. OpenAI says the accounts originated in Iran and the activity appeared consistent with a commercial, for-hire influence campaign, but it could not identify the actor.
Which outlets published the articles?
OpenAI does not name them. RFE/RL named PeaceVoice, CounterPunch and LA Progressive for one article, and CNN, as reported by VOA, named Middle East Monitor, which said it took the articles down and is reviewing its submissions process.
When were the articles published?
OpenAI says the earliest it found was from July 2025 and the latest from October 2026.
Should a vendor ever place an article under a made-up name?
No. An invented byline deceives the outlet and its readers. This is not legal advice.
Sources
This explainer is built from OpenAI's October 8, 2026 report, read through an Internet Archive copy of October 9, and two news reports on it, all read on October 17, 2026. Advice for contributors and buyers is ours, not a platform or outlet rule.
- platform documentation OpenAI, Disrupting AI-enabled 'false front' operations. Supports the bylines, counts, dates, prompts, attribution and Breakout Scale ratings. Read through an Internet Archive copy of October 9, 2026.
- trade press Radio Free Europe/Radio Liberty (Frud Bezhan), How Iranian Operatives Used AI To Publish Fake Stories In US Media. Supports the Hoskins article example and outlets. Read October 17, 2026.
- trade press VOA News, Open AI says Iranian regime used ChatGPT to create fake articles. Supports Middle East Monitor's statements to CNN and VOA's date range and outlet count. Read October 17, 2026.