OpenAI identified almost 100 articles under fake journalist identities

01

The identity opened the way for the articles

On October 8, OpenAI published an account of operations that, according to the company, used its models in campaigns linked to Russia and Iran. The stated goal was to insert geopolitical messages into outside audiences, beyond the channels controlled by the operators themselves. In the Iranian case, operators used seven identities presented as journalists and turned to the models to revise articles and pitches sent to online outlets.

The company says it identified almost a hundred articles published or distributed under these identities, in about a dozen small and midsize outlets. The report also describes a Russian operation that appears to have used an entity presented as a research center and contributors in Latin America. According to the assessment published by OpenAI, these people didn't know they were working for a Russian group.

Both cases, according to the document, combined models with traditional tactics. OpenAI notes that not all published content was produced by its models. The mechanism described also depended on identities, editorial pitches and publishing channels, a chain that simply reading the text doesn't reveal.

02

Publication doesn't prove impact

The report separates the circulation of the articles from activity on social media. OpenAI says the comments linked to the Iranian operation had little identifiable engagement. It also says the operators used misleading calculations to exaggerate the reach of their actions.

There is another important caveat about the Russian operation: part of the stories the operators claimed to have spread was not found or corroborated in open searches. OpenAI says it cross-checked internal reports with publicly available content and checks; some claims were corroborated, others weren't. So the numbers and attributions should be read as findings reported by the company, not as an independent audit of the whole campaign.

03

Checking starts before the text

For companies that receive outside analyses, pitches or research, fluency and a professional look aren't enough to confirm a source. A practical check can verify the author's identity and track record, the existence and funding of the institution, the original sources and the link between who publishes and who produced the material.

These steps don't guarantee the information is correct. They help avoid decisions based only on the form of the content and leave a record of who checked the evidence. On matters of reputation, investment or public communication, this chain of origin can matter as much as the factual review of the text.

Sources

OpenAI, 8 Oct 2026. https://openai.com/index/disrupting-ai-enabled-false-front-operations/