OpenAI banned a cluster of ChatGPT accounts used in a pro-Russia influence operation designed to criticize Western countries and promote a fabricated think tank, according to a company blog post. The network utilized virtual private networks to bypass restrictions on Russian users, deploying OpenAI’s technology to generate social media posts and articles that misattributed real academic research.
The International Burke Institute and Misattributed Research
The operation centered on the International Burke Institute, a website registered in February 2025 that describes itself as an “expert community” based in Israel. According to OpenAI, the site served as the core of the influence campaign. An analysis of the institute’s articles by the AI company revealed that 34 of 36 pieces were real academic studies written by other researchers but misattributed on the site. For instance, a legitimate migration governance article from the Migration Policy Institute was posted with a headshot and a false byline naming Kate Howell, a food science professor at the University of Melbourne. The website later changed the attribution to real author Kate Hooper after OpenAI published its findings.
Susan Ariel Aaronson, a professor of international affairs at George Washington University and co-principal investigator for the NSF-NIST Institute for Trustworthy AI in Law & Society, also found her work hosted on the site without permission. Aaronson told Fortune that the situation was disturbing, noting that it causes substantial distrust in a field where trust is critical for governance. The website’s listings also featured high-profile figures such as American linguist Noam Chomsky, political scientist Francis Fukuyama, Mike Pompeo, and Joseph Nye. Neither Chomsky nor Fukuyama responded immediately to requests for comment from Fortune. Other experts listed on the site, including Shlomo Avineri and Jiang Ping, passed away before the institute was founded.
Did you know? OpenAI’s investigation found linguistic clues indicating that drafts of the institute’s material were originally written in Slavic languages before translation. One article referred to Germany’s governing coalition as the “Svetofor coalition,” utilizing “svetofor,” the word for traffic light in several Slavic languages including Russian, according to OpenAI and reporting by TechRadar.
The Burke Sovereignty Index and Social Media Amplification
The website featured a “sovereignty index” designed to make Western countries appear inferior to Russia across political, economic, and military metrics, according to OpenAI. TechRadar noted that the index lacked consistent measurement standards, placing Vatican City above Spain and awarding Russia the highest military score. To promote these findings, operators utilized ChatGPT to generate social media captions and imagery, disseminating content across X, LinkedIn, Facebook, Substack, and Telegram.
While individual accounts on platforms like Facebook and X received low view counts, Telegram channels associated with the network attracted between 10,000 and 20,000 followers per channel. The Burke Institute doubled down on its legitimacy in a statement responding to French outlet Le Monde, asserting that third-party citations of its publications do not constitute evidence of coordination or a common purpose. Meanwhile, OpenAI reported disrupting another German-language operation earlier in the year that used ChatGPT to generate political content ahead of federal elections in Germany, criticizing the U.S. and NATO while supporting the far-right Alternative für Deutschland party.
Frequently Asked Questions
What was the International Burke Institute?
The International Burke Institute was a website registered in February 2025 that presented itself as an expert community based in Israel. According to OpenAI, it served as the core of a Russian influence operation that published misattributed academic articles and a favorable sovereignty index.
How did the operators bypass restrictions on Russian users?
The individuals behind the network used virtual private networks to circumvent OpenAI’s restrictions on Russian users before utilizing ChatGPT to generate social posts and translate content.
Were the scholars listed on the institute’s website actually involved?
No. Researchers such as Susan Ariel Aaronson confirmed they had no association with the institute, and OpenAI found that dozens of articles were stolen from actual authors and misattributed to different names or deceased figures to manufacture credibility.
Stay Updated on AI Security
Subscribe to our newsletter for the latest investigative reports on artificial intelligence, foreign influence operations, and digital governance.
Subscribe NowRelated reading