Foreign influence campaigns don’t necessarily need to build their own fake news sites anymore. In this case, an Iran-based operation got AI-generated articles into real publications.
In its report on the “Bogus Bylines” campaign, OpenAI says it identified nearly 100 articles published or syndicated across roughly a dozen outlets worldwide. The articles focused on the US-Iran conflict. OpenAI couldn’t identify the specific actor behind the operation.
AI turned influence work into a publishing pipeline
ChatGPT helped make each submission look like it belonged where it was going.
The operators asked it to review English-language drafts against a publication’s submission requirements and suggest edits. Once those changes were approved, they requested an email to pitch the editor. That made the process easier to repeat across outlets.
Seven fake journalist identities supplied the bylines. Some had supporting social accounts, and OpenAI says ChatGPT wrote one persona’s biography. Previous tests have also shown how easily chatbots can generate convincing fake news stories.
Together, those details gave editors something that resembled an ordinary contribution.
The real exploit was newsroom trust
The campaign worked through contributor channels publications already used.
The Washington Post reports that one article reached the Port St. Joe Star through PeaceVoice, an organization the paper had worked with for years. The publisher said he didn’t independently verify the writer’s identity or the article’s origin because of that existing relationship.
That familiarity gave the submission credibility before anyone checked who was behind it.
Some outlets later removed articles, and PeaceVoice tightened its AI policy, according to the Post. OpenAI banned the accounts involved, adding to its earlier efforts to disrupt malicious uses of ChatGPT.
What newsrooms have to verify now
AI detection won’t solve this by itself. The vulnerability includes the person supposedly behind the article.
Editors have more reason to verify contributors rather than accept a polished biography or familiar referral as proof of identity. Separately, research shows that AI chatbots can be manipulated into spreading misinformation.
OpenAI says it shared information about the campaign with relevant authorities. For editors, the practical takeaway is to check who submitted the piece and how it reached them, alongside checking the copy itself.