OpenAI Shuts Down Russian ChatGPT Accounts Behind IBI Influence Operation
OpenAI has disclosed that it banned a cluster of ChatGPT accounts operated from Russia that leveraged VPNs to circumvent geographic access restrictions and conduct a coordinated influence operation. The threat actors used the AI chatbot to generate and translate social media content published across Substack, Telegram, X, Facebook, and LinkedIn, while instructing the model to scrub linguistic markers that could expose their Russian origins. Researchers who flagged the activity noted that bypassing regional bans via VPN tunneling is a recurring tactic in state-sponsored campaigns; users concerned about exposure can validate their own connection using a VPN and proxy detection tool.
The operation centered on promoting the so-called International Burke Institute (IBI), hosted at ibi[.]institute, a site registered earlier this year that claims to be a global community of experts in economics, politics, and international relations. OpenAI's investigation found the campaign extended well beyond AI-generated posts, encompassing a website stocked with copied and misattributed academic papers, a custom "sovereignty index" scoring nations favorably toward Russia, and deliberate attempts to mask the operators' true geographic footprint. The domain's registration metadata, which can be inspected via a WHOIS lookup, provides one of the few concrete artifacts linking the infrastructure to its organizers.
The Burke Index, also branded as the Sovereignty Index, ranks countries across seven weighted categories: political, economic, technological, information, cultural, cognitive, and military sovereignty, each scored out of 100 for a 700-point maximum. Under this methodology, the United States tops the list at 650.9, followed closely by China at 649.1, Switzerland at 610.7, and Russia at 601.4. While OpenAI found no evidence the site articles were written by ChatGPT directly, many appeared to be plagiarized academic work with falsified attribution or drafted by a non-native English speaker and machine-translated before publication.
Despite the operation's relatively modest reach, its Telegram channels reportedly amassed between 10,000 and 20,000 followers each, and the bulk of generated content was published in English to maximize Western audiences. OpenAI emphasized that this campaign is distinct from previously documented Russia-linked influence operations, with its signature feature being the construction of a pseudo-academic index designed to legitimize pro-Kremlin narratives and criticize adversaries under the veneer of expert analysis. Readers can audit their own digital exposure to similar surveillance and data-collection risks by running a comprehensive privacy checkup.