Iranian and Russian networks used ChatGPT to spread political and military messages through media outlets and organizations that presented themselves as independent. This was reported by OpenAI, which blocked the accounts associated with the operations and passed the information to the relevant authorities.
„Most strikingly, these operations resemble sophisticated influence operations from the pre-AI era, but use AI to facilitate some of the work“, OpenAI says.
Seven fictional journalists and almost 100 articles
The Iranian campaign has been named Bogus Bylines. It used seven fake journalist identities, under whose names articles were offered to online publications. The authors were presented as Western journalists who write about international politics, the Middle East, human rights and conflicts.
OpenAI has identified almost 100 articles published or reprinted in about a dozen small and medium-sized media outlets around the world. The materials were mainly devoted to the conflict between the United States and Iran. The earliest discovered publication is from July 2025, and the most recent - from October 2026.
ChatGPT was not used only for writing. Operators fed the model English-language drafts and had it edit them according to the requirements of specific editorial offices. The system then prepared emails to send the texts to editors.
Some of the fictional journalists had profiles on social networks to appear credible. In one case, ChatGPT was also used to create a biography of the fictitious author. The campaign also included generating comments in English and Persian, but this activity has received limited attention.
Fictitious Research Center in Latin America
In parallel, OpenAI has shut down the Russian-linked operation Dark Clark. It operated through an organization called the Center for Social Research, which publicly presented itself as a research platform on the Indian diaspora in Latin America and relations between India and countries in the region.
According to OpenAI, Russian operators controlled the organization through a fake identity called Mia Clark. They recruited local employees who were likely unaware that they were working for a Russian operation. Some of them conducted interviews and prepared original analyses published by the center.
ChatGPT was primarily used to create and update internal reports. They described actions to undermine trust in Ukraine, influence politics in countries such as Argentina and Bolivia, and manage the fictitious organization.
In some cases, the model helped create fake documents, audio scripts, and texts tailored to the language and institutional style of a particular country. Some of the claims were related to Bolivia and Ecuador and prompted fact checks or official denials.
OpenAI rates the Russian operation at category five on a scale of one to six for the scale of influence operations. This is the highest category in which the company has placed an operation that has been interrupted since the beginning of its public reports. The Iranian campaign is rated at category four.
The company points out that the two networks did not rely solely on fake profiles on social networks. They have used seemingly independent media and organizations to reach audiences that would otherwise be difficult to attract through direct propaganda.
Sources: mezha.ua, therecord.media, channelnewsasia.com