OpenAI on Tuesday shut down a network of Russian ChatGPT accounts that used VPNs to evade access restrictions and power an influence operation across multiple social media platforms. The accounts leveraged the company's AI tool to generate posts and comments that were distributed on Substack, Telegram, X, Facebook, and LinkedIn. The campaign centered on promoting the International Burke Institute, described by the organization as an expert community based in Israel.
The banned accounts instructed ChatGPT to generate content primarily in English while concealing any language patterns that might reveal their Russian origins, according to OpenAI. The operation reached relatively modest audiences, with Telegram channels drawing the most engagement at roughly 10,000 to 20,000 followers each. The IBI website was registered in February 2025 and presents itself as a global community of leading experts in economics, politics, sociology, and international relations. While OpenAI found no evidence that articles on the site were created using ChatGPT, many appeared to be copied from academic writings—sometimes with false attribution—while others seemed to be written by a Slavic speaker and then machine-translated into English.
The company said what made this operation distinctive was the creation of a sovereignty index promoted by the think tank to portray Russia favorably while criticizing Western nations. The Burke Index calculates sovereignty across seven indicators—political, economic, technological, information, cultural, cognitive, and military—with each area scored out of 100 points based on official national and global data sources plus expert assessments, yielding a final cumulative score ranging from 100 to 700. The IBI website's sovereignty rankings placed the United States first at 650.9, followed by China at 649.1, Switzerland at 610.7, and Russia at 601.4. The report on Russia characterized the country as "a state with a high level of sovereignty" and "one of the three world powers with full digital and military nuclear sovereignty," while noting challenges including demographic aging and innovation constraints. Beyond IBI content generation, one ChatGPT user created a profile picture for a Telegram channel called "American Observer" and repeatedly requested Russian-language summaries of the channel's activity.
OpenAI emphasized that the operation's significance lies more in the infrastructure it constructed than in the audience it reached. While the actors only used ChatGPT to produce isolated promotional posts, those posts pointed to what appeared to be a credible institution equipped with purported experts, republished academic work, and a proprietary risk index. The company explained this case illustrates how influence actors can deploy AI as a supporting tool within a broader effort to manufacture authority, obscure the source of preferred narratives, and establish assets that could be scaled over time—and how their supporting use of AI can lead to the broader operation being exposed. The takeaway is clear: AI tools are becoming infrastructure for disinformation campaigns that build legitimacy over months before scaling, not just content generators for viral posts. For security teams and platform operators, the real threat isn't the AI-generated text itself but the credible institutional facades that text helps construct and maintain.