OpenAI has banned a cluster of ChatGPT accounts that very likely originated in Russia after detecting their use to promote the International Burke Institute (IBI), an organization presented online as an Israel‑based “expert community.” What began as an inquiry into AI-generated social media posts expanded into the discovery of a broader influence operation built around a website (registered in February 2025), republished or misattributed academic material, a pro‑Russia “sovereignty index,” and efforts to conceal the operators’ Russian origins.
How the operation worked
The accounts used Russian-language prompts to generate social media comments that were then posted on Substack, Telegram, X, Facebook and LinkedIn. Although operators prompted the model in Russian, most generated content was in English, and operators specifically instructed ChatGPT to remove linguistic cues of Russian origin. Because OpenAI does not permit access to its models from Russia, the operators used VPNs to reach the platform.
A substantial portion of the generated content promoted the International Burke Institute brand. The IBI website — registered in February 2025 and claiming an Israeli address — published a range of articles and presented purported experts. OpenAI’s review found the site’s articles were not produced by OpenAI models; instead, many pieces were copied from other internet sources, sometimes attributed to the wrong authors, and in several cases showed signs of being drafted by a Slavic speaker and then machine translated. One cited example on the site referred to Germany’s coalition as the “Svetofor coalition,” a literal translation from Slavic languages for “traffic light,” which is not natural in English or German and likely indicates machine translation from a Slavic original.
Republished and misattributed academic content
OpenAI examined a sample of 36 articles linked to experts on the IBI website, published between September 2025 and May 2026; 34 of those articles were found to be copied from elsewhere online. Some pieces came from years-old sources; others were incorrectly attributed. For instance, an article about the China–Pakistan Economic Corridor appears to have been taken from a Cambridge University Press original but was misattributed to a University of Nottingham professor whose expertise lies in South Asian politics. Another migration‑governance article seems copied from the Migration Policy Institute but attributed to an Australian food‑science professor.
The IBI website also published reports based on a proprietary‑style “sovereignty” index that praised Russia and criticized Western countries. These reports compared countries or focused on single countries perceived as critical of Russia’s war in Ukraine, such as France, Germany and the European Union, and similar critical narratives were promoted about the United States.
Social media activity and supporting assets
ChatGPT’s primary role in the operation was to generate social posts promoting IBI website articles. OpenAI identified such posts across multiple platforms; some were posted by accounts bearing the IBI name and logo, while others appeared to be inauthentic ordinary users whose activity mainly consisted of posting IBI content. The operators also generated replies to real Substack users across a range of topics; these replies typically invited readers to follow the IBI channel.
One operator produced German‑language posts published on a Telegram channel called “Lahme Ente” ("lame duck") that regularly criticized Ukraine, the EU and the German government and advocated warmer relations with Russia. A second operator created logos for about a dozen Telegram channels focused on Germany, the USA, France, Poland and Türkiye; some of those channels occasionally posted or shared IBI material. This second operator also requested Russian‑language summaries of those channels’ activity. One U.S.-focused channel presented itself as an American outlet but used non‑idiomatic English in its profile, suggesting non-native authorship.
People and provenance
Although the operation’s centerpiece appeared to be the International Burke Institute, the disrupted activity mainly consisted of social media promotion and linking to the website rather than content hosted and produced on the site itself. OpenAI’s investigation found online traces that suggest a handful of real individuals in Israel may have acted on behalf of the IBI at conferences or in article submissions, but OpenAI did not establish a definitive relationship between those individuals, the IBI website, and the Russian operators.
Reach and assessment
The immediate impact of the campaign was limited: typical social posts had low view counts and official IBI accounts had few subscribers. Telegram channels tied to the operation attracted larger followings, generally about 10,000–20,000 followers each. Using the Brookings Breakout Scale to evaluate impact, OpenAI placed the operation at the lower end of Category Three — multiple platforms were used and there were some signs of breakout into authentic audiences.
The operation’s significance comes more from the infrastructure it assembled than its measured reach. Operators used ChatGPT mainly for isolated promotional posts, but those posts pointed to a seemingly credible institution complete with purported experts, republished academic work and a proprietary index. The case illustrates how influence actors can use AI as a supporting tool within a broader effort to manufacture authority, obscure the origin of narratives, and build assets that could be scaled over time — and how the supporting use of AI can itself leave traces that help expose the broader operation.
