Eigen RadarAI
Analysis

OpenAI discloses influence operations using fictitious writers in existing media

OpenAI has banned accounts it links to Russian and Iranian influence operations that used fictitious identities to place material in existing media. Its findings describe activity aimed at Latin America and articles about the US–Iran conflict. NPR also obtained responses from affected publishers. Attribution and reach estimates remain OpenAI’s assessments, including its finding that some operators exaggerated their accomplishments in internal reports.

Artificial Intelligence··Evening
A publication workstation with a monitor back, keyboard, mouse and empty editor chair.

Fictitious identities placed material in existing outlets

OpenAI disclosed findings about Russian and Iranian accounts using its tools for covert influence operations and said it banned them. The company calls the operations Dark Clark and Bogus Bylines. According to its findings, both used fictitious identities to place material in established media outlets. Their attribution to Russian and Iranian actors remains OpenAI’s assessment.[1], [2]

Latin American targets and nearly 100 articles feature in the findings

OpenAI says Dark Clark targeted Latin America to weaken support for Ukraine and influence local politics. Fabricated audio and documents prompted fact-checks and denials in Ecuador and Peru, according to the company. OpenAI also found that users prepared internal reports for supervisors with ChatGPT and that some of those reports exaggerated what the operators had accomplished.[1], [2]

The company says Bogus Bylines used seven fictitious writers to place nearly 100 articles, including material about the US–Iran conflict. Its social-media comments gained little traction in OpenAI’s account. AI also helped adapt material into different languages. These descriptions and the overall reach estimates are the company’s findings.[1], [2]

Publishers describe their submission safeguards

NPR obtained responses from publishers beyond OpenAI’s account. Daily Kos said its community section is separate from staff journalism and allows user posts without prior staff review. Middle East Monitor said it was reviewing submissions and strengthening editorial safeguards. A researcher at Clemson University described how AI can make foreign-language material appear authentic. The publisher responses describe how submissions are handled without independently settling every claim about the operations.[2]

References

  1. News sourceThe DecoderOpenAI discloses Russian and Iranian influence campaigns↩1↩2↩3
  2. News sourceNPROpenAI discloses Russian and Iranian influence campaigns↩1↩2↩3↩4