Pull to refresh
Logo
What Changed
Following Why Sign Up
Iranian operatives used ChatGPT to plant AI-written articles in U.S. outlets

Iranian operatives used ChatGPT to plant AI-written articles in U.S. outlets

New Capabilities

OpenAI says fake personas slipped nearly 100 opinion pieces past real editors

October 9th, 2026: Washington Post details the fake bylines

Overview

Updated Oct 20

A Florida weekly published an op-ed on the U.S. war with Iran under the byline Ervin Hoskins, a writer who does not exist. OpenAI says Iranian operatives created him and six other personas with ChatGPT to plant nearly 100 articles in real news outlets.

OpenAI rated the effort Category 4 of 6 on its influence-operations scale and said it could not identify the commercial actor behind it. PeaceVoice, the distributor that unwittingly pitched the fake column, has since strengthened its no-AI policy. Daily Kos and Middle East Monitor removed the articles, and Port St. Joe Star publisher David Adlerstein said he is reviewing an explicit AI policy.

Why it matters

Foreign operatives can now place AI-written opinion pieces in genuine outlets, reaching readers who never learn the byline is fake.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

~100
Fake articles placed
AI-generated or AI-polished op-eds published or syndicated under the operation's bylines, per OpenAI.
7
Fake journalist bylines
Fake personas with social media profiles and AI-generated images.
20+
Outlets that ran them
Washington Post count; OpenAI said roughly a dozen online outlets.
4 of 6
Score on OpenAI's influence scale
Category 4 on the six-level IO Breakout Scale, which tracks campaign impact.

Voices

Curated perspectives — historical figures and your fellow readers.

Sign up to generate historical perspectives on this story.

Log in Sign Up

People Involved

Ervin Hoskins
Ervin Hoskins
Fabricated journalist byline used by Iranian operatives
Exposed as fake by OpenAI's October 2026 report
 David Adlerstein
David Adlerstein
Publisher, Port St. Joe Star
Says he is examining an explicit AI policy
 Markos Moulitsas
Markos Moulitsas
Founder of Daily Kos
Condemned the campaign and is examining prevention
JB
Jessica Brandt
Former director, Foreign Malign Influence Center, Office of the Director of National Intelligence (2023–2025)
Commented publicly on the campaign

Organizations Involved

OpenAI
OpenAI
Artificial intelligence company
Disclosed the campaign and banned the accounts behind it

Builds ChatGPT; publishes threat research on misuse of its models.
PeaceVoice (Oregon Peace Institute)
PeaceVoice (Oregon Peace Institute)
Nonprofit op-ed distribution service
Strengthened its no-AI policy after OpenAI contact

Distributes opinion pieces to newspapers; run by the Oregon Peace Institute.
Middle East Monitor
Middle East Monitor
London-based news and analysis outlet
Published at least 21 articles under fake bylines; later removed author pages

UK-based outlet covering the Middle East; brands itself supportive of the Palestinian cause.
Daily Kos
Daily Kos
Progressive political news and community site
Removed articles by fake bylines; founder examining prevention

Community section lets any registered user post without staff review before publication.

Timeline

June 2026 October 2026

4 events Latest: October 9th, 2026 · 1 month ago
Tap a bar to jump to that date

  1. Washington Post details the fake bylines

    Latest Investigation

    The Post reports the Port St. Joe Star column was AI-generated; affected outlets take down articles and author pages.

  2. PeaceVoice strengthens no-AI policy

    Policy change

    PeaceVoice, the op-ed distributor that unwittingly pitched the fake Hoskins column, strengthened its policy against AI-generated submissions after OpenAI contacted it about the case.

  3. OpenAI discloses Iranian influence campaign

    Disclosure

    OpenAI publishes a report linking seven fake bylines to Iranian users, bans their ChatGPT accounts, and blocks access from Iran.

  4. Fake byline article enters circulation

    Influence operation

    PeaceVoice runs Ervin Hoskins' column on U.S. Iran policy; other outlets republish it under similar headlines.

Scenarios

1

OpenAI names the operator behind the Iranian campaign

Unlikely Resolves by Apr 9, 2027

Discussed by: OpenAI's report; security researchers including former ODNI official Jessica Brandt

OpenAI said the operation looked like a commercial for-hire effort but that it could not identify the actor. If law enforcement or an investigative outlet names the firm or group behind it, the case moves from anonymous to attributable. That would require a follow-up disclosure, a court filing, or an investigation identifying a specific entity.

2

Small newsrooms adopt AI-content rules after the hoax

Possible Resolves by End of 2026

Discussed by: Port St. Joe Star publisher David Adlerstein; Daily Kos founder Markos Moulitsas

Adlerstein said he is examining an explicit AI policy; Moulitsas said he is looking into prevention. If affected outlets — the Star, Daily Kos, Middle East Monitor — publicly announce written policies on AI-generated content or on verifying bylines, the campaign's edge case becomes an institutionalized lesson.

3

New AI influence campaigns surface despite the disclosure

Likely Resolves by Oct 1, 2027

Discussed by: Security researchers tracking influence operations

The economics that made this campaign work — cheap, fluent AI copy that reads like a freelance writer — are not going away. OpenAI, Meta, or academic researchers may disclose a new AI-assisted campaign placing content in genuine outlets, either Iranian or copying the for-hire model.

Historical Context

2 moments from history that rhyme with this story — and how they unfolded.

2013–2016

Internet Research Agency (IRA) (2013–2016)

A Russian troll farm in St. Petersburg ran hundreds of fake American personas across Facebook, Twitter, and YouTube, buying ads and organizing around the 2016 U.S. election. The U.S. Justice Department indicted the IRA and its owner Yevgeny Prigozhin in 2018.

Then

Facebook and Twitter removed accounts; Congress subpoenaed tech executives but passed no targeted law.

Now

Platforms built election-integrity teams and ad-transparency tools, and the IRA became the template for foreign influence operations.

Why this matters now

The playbook — fake Western personas pushing polarized content — now runs on AI generation, removing the need for a large human staff writing fluent English.

June–August 2018

Facebook removes Iranian influence networks (2018)

Facebook in mid-2018 took down coordinated inauthentic behavior linked to Iran, removing pages and accounts that posted about U.S. politics and the Middle East. Twitter and other platforms announced similar takedowns.

Then

The networks were removed and Iran's capabilities became public.

Now

Iran remained a recurring actor in U.S.-targeted influence; platforms added takedown playbooks for state-linked networks.

Why this matters now

Iran's targeting of American audiences predates ChatGPT. What is new here is scale and polish, not the willingness to impersonate American writers.

Sources

(11)
+5
The Washington Post NPR Radio Free Europe/Radio Liberty The Verge WQOW (CNN) Iran Times (Washington Post syndication) Attention to the Unseen DiggingBeagle PNP Iran International The Plain Signal