Russia's bot network used ChatGPT in a massive covert operation
Russian bots used ChatGPT to influence the EU (Photo: Unsplash)
OpenAI has identified and blocked a network of ChatGPT accounts operated from Russia that promoted a fake Israeli think tank and spread Kremlin narratives.
How the network worked
The Russian operators used ChatGPT through VPNs and entered prompts in Russian.
They instructed the AI to generate comments in English and German while hiding any linguistic traces of Russian origin.
The AI-generated content was then posted on X, LinkedIn, Facebook, Substack, and Telegram. The main goal was to promote the International Burke Institute (IBI) website.
The operators also used AI to:
- Write comments under posts by real Substack users encouraging them to subscribe to IBI.
- Create German-language posts for the Lahme Ente Telegram channel, criticizing the German government, the EU, and Ukraine.
- Generate logos and profiles for dozens of Telegram channels targeting audiences in the United States, France, Poland, and Türkiye.
Fake institute and stolen articles
The core of the operation was the IBI website, which was registered in February 2025 with an address in Israel. The platform claimed to work with leading scholars, including Noam Chomsky and Francis Fukuyama.
But an investigation uncovered widespread fabrication.
What exactly was fake
Plagiarism: Thirty-four of the 36 articles analyzed were copied in full from legitimate academic sources, including Cambridge University Press and the Migration Policy Institute.
Fake authorship: The articles were attributed to unrelated scholars. In one case, an article about migration was published under the name of an Australian professor specializing in food chemistry.
Machine translation: The texts contained unnatural phrasing. For example, Germany's "Ampelkoalition", or "traffic light coalition" was rendered in English as "svetofor coalition," suggesting the phrase had been translated literally from Russian.
The operation’s main tool for promoting its manipulation was a fabricated "sovereignty index." In these reports, the Russian operators praised Russia's autonomy while harshly criticizing France, the United States, Germany, and the European Union.
Assessing the campaign’s impact
According to OpenAI, the social media posts had limited direct reach. However, the Telegram channels created as part of the operation attracted between 10,000 and 20,000 followers each.
On the Brookings Breakout Scale, the operation received a score of 3, indicating that it had spread across multiple platforms with occasional reach into real audiences.
OpenAI said the operation highlights a new trend: using AI not only to generate fake content but also to manufacture false authority and imitate legitimate analytical institutions.
