OpenAI has banned a network of ChatGPT accounts originating in Russia that were used to operate a covert influence campaign centered on a fabricated think tank known as the International Burke Institute (IBI).
According to a threat intelligence report published by OpenAI on August 25, 2026, the operation used ChatGPT to generate English-language social media content across platforms including X, LinkedIn, Facebook, Substack, and Telegram. The operators prompted the models in Russian while explicitly instructing them to conceal linguistic markers of Russian authorship and circumventing access restrictions via VPN connections.
Structure of the Burke Institute Operation
The influence campaign was constructed around the IBI website, registered in February 2025 with a claimed physical address in Israel. The site presented itself as an independent expert community, listing high-profile academics and publishing extensive policy analyses.
OpenAI's investigation revealed that the underlying content was largely synthetic or plagiarized. In an analysis of 36 articles published on the institute's site between September 2025 and May 2026, researchers found that 34 articles were copied from existing publications without authorization, frequently with altered or falsified author attributions. For instance, a paper originally published by Cambridge University Press was attributed to a University of Nottingham professor, while an analysis from the Migration Policy Institute was credited to an Australian food science academic.
The core ideological vehicle on the site was a proprietary "sovereignty index," which ranked global governments using criteria designed to score Russia favorably while rating Western European democracies poorly.
Multi-Platform Distribution and Channel Networks
ChatGPT was primarily utilized at the distribution layer of the campaign to generate promotional commentary and replies directing traffic to the IBI repository. Beyond direct think tank promotion, operators used the model to maintain localized regional channels:
- German-Language Channels: One account managed the Telegram channel "Lahme Ente" (Lame Duck), generating posts focused on criticizing the German government, European Union policy, and military support for Ukraine.
- Targeted Outlets: A second operator generated channel branding and logos for roughly a dozen regional Telegram outlets aimed at France, Poland, Turkey, and the United States, including a persona named "American Observer."
- Audience Engagement: Operators used ChatGPT to draft contextual replies on Substack to solicit followers and boost engagement on authentic posts.
Threat Assessment and Reach
OpenAI evaluated the operation at Category Three on the Brookings Breakout Scale, noting cross-platform deployment and modest penetration into authentic audiences. While typical social media posts registered low engagement numbers, several of the regional Telegram channels maintained subscriber bases between 10,000 and 20,000 followers.
The incident highlights a continuing pattern in adversarial AI use: deploying large language models not to synthesize core disinformation from scratch, but rather to automate translation, disguise operator origin, and scale distribution across authentic discussion forums.
