OpenAI reports influence operations hiding behind bylines and a research front

OpenAI reports one operation using seven journalist personas and another controlling a Latin American research front whose staff appeared unaware of its sponsors. The company’s attribution remains its investigative assessment. [1]

OpenAI identified almost 100 published articles under the Iran-origin operation’s bylines across more than a dozen publications. That is a distribution finding, not a count of independent investigations or people convinced. The Russia-origin cluster used ChatGPT heavily for internal reporting; the front’s staff produced much of its original research. AI assistance should not be confused with AI authoring everything. [1]

Independent checks of public material can expose the disguise. Ecuador’s Lupa Media had already examined a fake report attributed to Teleamazonas, finding no matching broadcast and photographs taken from unrelated Instagram accounts. Its work corroborates that fabrication, not OpenAI’s private account telemetry or the complete country attribution. [2]

For this publication and its readers, the useful response is to inspect the chain behind a claim: who gathered the evidence, who owns the organisation and whether repeated articles share the same origin. That is our editorial recommendation. A familiar masthead or fluent writing makes none of those checks unnecessary.