OpenAI reports disrupting influence operations using false public identities

False identities supported by AI

OpenAI says it banned two covert influence operations that used its models alongside more traditional techniques. Its October 8 report attributes one operation to Russia and the other to Iran, describing efforts to distribute geopolitical messaging through misleading public identities.

The company says the Iran-linked operation used seven journalist personas to pitch articles to outside publications. It describes the Russia-linked operation as apparently using unwitting people in Latin America to operate a think tank. These are OpenAI’s findings and attributions, rather than an independent investigation conducted by Franklin.

Look at the distribution path

OpenAI reports that both operations used AI for internal reporting as well as other tasks. That detail broadens the question beyond whether a model wrote the final article. An operation can use AI to organize work or present its own performance even when other people produce some of the material that reaches an audience.

For an editor, the useful response is to examine the source of a pitch and the identity behind it. A fluent article should still have verifiable authorship, supporting evidence and a clear account of any organization represented by the writer. A plausible biography does not settle those questions.

Preserve evidence behind a decision

The report illustrates why publication channels deserve attention alongside generated text. A decision to accept outside material gives the claimed author access to an established audience and the publication’s reputation.

Editors reviewing unfamiliar contributors can retain the pitch, the stated affiliation and the evidence used to check it. If a concern emerges later, those records make it easier to explain why material was accepted and what needs correction. That is a workflow lesson drawn from the reported cases, not a claim that ordinary contributors are deceptive or that an automated authorship detector can establish who controls a persona.