OpenAI says it stopped Russian and Iranian groups using ChatGPT for propaganda

What OpenAI found

OpenAI says it has banned accounts linked to Russian and Iranian influence operations that used ChatGPT to support propaganda efforts. According to Recorded Future News↗ (opens in a new tab), these groups used AI tools to help create fake research organisations, generate articles, and pitch stories to media outlets. The company says the Russian campaign focused on Latin America, while the Iranian campaign targeted international news coverage of US-Iran relations.

How the Russian operation worked

The Russian-linked group set up a fake research organisation called the Social Research Center (SRC). It claimed to study the Indian diaspora in Latin America and regional relations, but OpenAI says it was controlled by Russian operatives using a false identity. The group recruited local researchers, who apparently did not know about the Russian connection, to produce original research. ChatGPT was used to create fake leaked documents and scripts for audio recordings. Some of these materials spread online and were picked up by schools, a politician, and media outlets, according to NBC News↗ (opens in a new tab). Some claims were circulated widely enough to prompt fact-checks and official denials.

The campaign mainly aimed to damage Ukraine’s reputation in Latin America, but also tried to influence domestic politics in countries such as Argentina and Bolivia. OpenAI did not name a specific group behind the campaign but said it may be linked to Politology or La Compania, which have been reported as successors to operations associated with the late Russian oligarch Yevgeny Prigozhin and the Wagner Group.

The Iranian campaign

The Iranian operation used seven fake journalist identities to pitch articles about US-Iran conflict to news outlets. According to France 24↗ (opens in a new tab), these profiles posed as Western journalists and managed to place nearly 100 articles in about a dozen publications. Most of these were small or midsize outlets focused on international affairs and the Middle East. Some of the outlets had large audiences, with one having nearly two million Facebook followers. The operation used ChatGPT to improve articles, prepare pitches, and generate social media comments in English and Persian. However, OpenAI says the social media posts attracted little attention.

How AI was used

OpenAI’s report highlights that these operations blended traditional influence tactics with new AI tools. ChatGPT was used to speed up tasks like drafting documents, improving articles, and generating scripts. The operators accessed ChatGPT from Russia and Iran using VPNs, since OpenAI does not allow direct access from those countries. OpenAI says the overall approach resembled older influence operations, but AI made some parts of the workflow easier and faster.

What happens next

OpenAI has shut down the accounts involved and published its findings. The company did not identify the individuals or organisations responsible for the Iranian campaign, nor did it name the media outlets that published the articles. There is no public evidence yet from government agencies or independent researchers confirming the details, as all information comes from OpenAI’s own report. The story shows how AI tools can be misused for information operations, and why monitoring and response are needed as these technologies spread.

What you should do

For organisations and individuals, this case is a reminder to question the source of online content, especially when it claims to be leaked or comes from unfamiliar research groups or journalists. Media outlets should be cautious about accepting pitches from unknown contributors. Security teams should watch for signs of coordinated influence activity, such as repeated narratives or suspicious new organisations. For AI providers, monitoring for misuse and enforcing access controls remain critical.