OpenAI: fake AI journalists placed ~100 articles. Verify bylines
OpenAI says seven AI-assisted fake journalist personas placed almost 100 articles in real outlets. If you accept guest posts or press pitches, verify first.
Fake AI-assisted journalists got almost 100 articles into real publications. In an October 8 report, OpenAI said it banned two covert influence operations that used ChatGPT to support "false front" identities. One ran seven invented Western journalist personas that pitched articles to editors. The other ran a think tank through a fake persona. This was geopolitical propaganda, but the method works on any business that publishes outside writing, accepts guest posts or answers press requests.
What actually happened
Per CyberScoop and Euronews, summarizing OpenAI's report:
- The Iran-origin operation used at least seven fake journalist profiles, each claiming to be from a Western country. OpenAI found "almost 100 articles published or syndicated under the operation's bylines across roughly a dozen online outlets," mostly small and medium sites on international affairs. The earliest dated to July 2025. OpenAI called it a commercial, for-hire campaign.
- The Russia-origin operation ran the Social Research Center, a supposed Latin American think tank fronted by an AI persona named "Mia Clark." OpenAI said its local staff "were not aware that they were working for a Russian group."
- Ratings: On OpenAI's six-point severity scale, the Russian campaign rated 5 and the Iranian one 4. CyberScoop notes most campaigns rate 1 or 2.
One Euronews detail stands out: a persona that called itself "an American freelance writer" had an X account that platform transparency data linked to a "West Asia android app," and an Instagram account listed as based in Iran. The signal was there. Nobody checked it.
Why it matters for your business
Small outlets were the target because their review process is thin. So is yours. If your site accepts guest posts, your newsletter runs contributor pieces, or you answer "journalist" emails asking for a quote, you are running an intake process with no identity check. AI makes a convincing pitch, writing sample and backstory almost free. Expect more of them, and not only from states. Link-buyers, competitors and scammers use the same tools.
A cheap verification checklist:
- Check the account history. Profile age, location data, and published work that predates the pitch.
- Get on a live video call before you publish anyone new. Seven fake reporters cannot all show up.
- Check outbound links in submitted pieces. Many guest-post schemes exist only to plant them.
- Log who approved what. If a piece turns out to be fake, you need to find the rest of that writer's work fast.
Key takeaways
- OpenAI disclosed two influence operations on Oct. 8 that used ChatGPT to support fake identities
- Seven fake journalist personas placed almost 100 articles across about a dozen real outlets
- A Russian operation ran a think tank through an AI persona; local staff did not know who they worked for
- Small publishers were targeted because their checks are thin
- Verify contributors with account history, a live call and a link review before you publish
Accepting submissions, reviews or vendor applications through a form? We build intake workflows with verification steps, approval logs and AI-assisted screening that flags a pitch before it hits your site. See what we build or tell us about your intake process.
Sources: OpenAI, CyberScoop, Euronews.
- #openai
- #fake-journalists
- #influence-operations
- #content-verification
- #guest-posts
Tommy Rush — Founder, Rush Commerce
Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More
Get The Rush Report weekly — one email, zero fluff.
Keep reading
White House: AI incident reporting is 'not optional' now
After Anthropic's agent incidents, the White House says AI incident reporting is not optional. No penalties yet. Log your own agents' actions today.
Read itAI agent trust falls from 75% to 56%: keep human review
VB Intelligence finds trust in AI agents shipping production changes unreviewed fell from 75% to 56%. Why human review is the right default for small teams.
Read it