OpenAI Bans Russian ChatGPT Network Behind Fake Burke Institute Think Tank

OpenAI announced it has banned a cluster of ChatGPT accounts linked to a Russian influence operation that spent over a year building credibility for a fictitious think tank called the International Burke Institute. The operators connected via VPN to bypass the service's unavailability in Russia, issuing prompts to ChatGPT in Russian while requesting English-language commentary and articles that were then published on Substack, Telegram, X, Facebook, and LinkedIn.

How the fake institute operated

The International Burke Institute presented itself as an independent research center based in Israel, publishing geopolitical and economic analysis. In reality, most of the text on its site was copied from other online sources and credited to people who had nothing to do with it. One article on the China-Pakistan Economic Corridor originally came from Cambridge University Press but was attributed to a University of Nottingham lecturer who never wrote it.

Another example is a piece on migration management actually written by Kate Hooper of the Migration Policy Institute, wrongly credited to Kate Howell, a food science professor unconnected to the topic. The institute's list of purported experts also included well-known names from academia, including political scientist Francis Fukuyama and linguist Noam Chomsky, neither of whom ever worked with IBI.

A sovereignty index and the campaign's goal

At the center of the operation was the so-called Sovereignty Index, also known as the Burke Index, a ranking of countries by their political independence. According to OpenAI, the index consistently cast Russia in a favorable light while criticizing Western states. Content promoting the ranking and the institute itself was posted across multiple social platforms simultaneously, aiming to create the impression of an independent, widely cited source.

The operation was also linked to a German-language Telegram channel called Lahme Ente, which published content attacking Ukraine and the European Union. OpenAI assesses that despite the elaborate infrastructure, the campaign's reach stayed limited - most social media posts drew only small numbers of views, and IBI's official accounts had few subscribers.

Scholars react to impersonation

Some of the people whose names appeared on the institute's site without their knowledge spoke out publicly. Susan Ariel Aaronson, a George Washington University professor listed on IBI's site, described the situation as damaging trust in artificial intelligence itself.

Trust is really important in AI governance, and here we have a really strange situation that creates serious distrust - Susan Ariel Aaronson, professor at George Washington University

Another chapter in Russian influence operations

This is not the first time OpenAI has exposed a Russian influence operation using ChatGPT. In June 2025, the company described a separate campaign producing German-language political content promoting the Alternative for Germany party ahead of Germany's federal elections. The company notes, however, that the Burke Institute operation is by far the most elaborate and previously undocumented structure of its kind that it has managed to shut down.

OpenAI regularly publishes reports on detected abuse of its models by state actors, having previously disclosed attempts by Chinese and Iranian groups to use ChatGPT for producing propaganda and analyzing malware code. Such disclosures are meant to show that the company actively monitors abuse, but they also confirm the scale of attempts to use generative AI for automated disinformation.

For users and institutions outside Russia, the case is a reminder that content published under the banner of seemingly independent think tanks requires extra source verification, especially when it invokes well-known names. Increasingly easy access to generative tools lowers the cost of building such facade institutions, and uncovering the fraud often depends on someone whose work was copied happening to notice the plagiarism.