A fake institute presented as a group of experts based in Israel, publications favorable to the Kremlin and a use of ChatGPT to hide the Russian origin of their authors: OpenAI claims to have dismantled an influence operation which used its artificial intelligence to produce and distribute content on several platforms.
OpenAI’s investigation began after the discovery of social media posts generated by artificial intelligence. She led the company into a larger operation organized around the International Burke Institute (IBI), a purported Israel-based think tank that OpenAI says doesn’t actually exist.
The fake institute, however, had its own website, registered in February 2025. It displayed the names of known specialists and published alleged scientific work that OpenAI describes as copied and wrongly attributed. The IBI claimed in particular to bring together leading international experts. The IBI claimed in particular to bring together leading international experts. It also published a “sovereignty index” that looked critically at Germany, France, the European Union and the United States, while rating Russia positively.
Distribution on social networks
Those responsible for the operation used ChatGPT to write comments and posts favorable to the Kremlin, then distributed on X, Facebook, LinkedIn, Substack and Telegram. OpenAI claims they were connecting via VPN to circumvent the blocking of its services in Russia. They also asked the chatbot to remove linguistic clues that might betray their Russian origin. A significant part of the content produced was used to give visibility and credibility to the fake institute.
OpenAI indicates that it has suspended several ChatGPT accounts linked to Russia and notified the platforms concerned. Several accounts on social networks were then deleted.
The mechanisms of disinformation exposed
Above all, OpenAI’s investigation revealed the workings of the construction of a propaganda operation: creation of a false institute to give it the appearance of legitimacy, usurpation of the authority of recognized experts, production of content by AI and concealment of their Russian origin. An elaborate device, but whose audience ultimately remained limited.
According to OpenAI, most posts generated few views and the official IBI accounts had few followers. Only its Telegram channels have had more success, with generally 10,000 to 20,000 subscribers. Ben Nimmo, responsible at OpenAI for tracking influence operations, summarizes with “Spiegel”: “advanced tactics”, but sometimes clumsy implementation.
