OpenAI Bans Russian ChatGPT Accounts Behind Covert Influence Operation Targeting Western Audiences
The operation used AI-generated content to promote a fake Israel-based think tank, complete with a proprietary “Sovereignty Index” designed to cast Russia in a favorable light while criticizing Western nations.
OpenAI announced Tuesday that it has banned a cluster of Russian ChatGPT accounts that employed virtual private networks (VPNs) to circumvent access restrictions and conduct a sophisticated influence operation. The accounts were used to generate social media posts and comments that were subsequently distributed across multiple platforms, including Substack, Telegram, X, Facebook, and LinkedIn. According to the company’s threat intelligence team, the accounts were being used to promote the International Burke Institute (IBI), a self-described “expert community” ostensibly based in Israel. What began as a routine investigation into AI-generated social media posts quickly unraveled into something far more complex: a broad influence operation built around a website containing copied and misattributed academic work, a fabricated “sovereignty” index that portrayed Russia in a favorable light, and deliberate efforts to disguise the operators’ Russian origins. The campaign, which OpenAI assessed to have reached “relatively small audiences,” nonetheless represents a significant evolution in how state-aligned actors are leveraging artificial intelligence to manufacture credibility and influence public discourse in the digital age.
The International Burke Institute, the centerpiece of this operation, presents itself as a legitimate community of leading experts spanning economics, politics, sociology, and international relations. Its website, registered in February 2025, projects an image of academic respectability and global expertise. However, OpenAI’s investigation revealed that the institution’s credibility was entirely fabricated. While there is no evidence that articles published on the site were created using ChatGPT, many of them appear to have been copied directly from existing academic writings, in some cases with false attribution to individuals who had no connection to the work. Other articles were seemingly drafted by a Slavic speaker and then machine-translated into English, leaving subtle linguistic traces that hinted at the operation’s true origins. The IBI brand was carefully constructed to appear legitimate, complete with purported experts and republished academic work, creating what OpenAI described as an “otherwise credible-appearing institution.” This infrastructure was designed to obscure the source of favored narratives and establish assets that could potentially be scaled over time, demonstrating how influence actors can use AI as a supporting tool within a broader effort to manufacture authority and deceive unsuspecting audiences.
The most distinctive element of the operation was the creation of a “Sovereignty Index,” also referred to as the Burke Index, which the think tank used to project Russia favorably and criticize Western countries. The index is calculated using seven indicators spanning political, economic, technological, information, cultural, cognitive, and military sovereignty. According to the IBI website, each of the seven areas is calculated using equalization coefficients based on a maximum score of 100 points, derived from official national and global data sources and expert assessments. The sum of the seven indices, ranging from 100 to 700, forms the final Cumulative Sovereignty Index. The rankings place the United States at the top with a score of 650.9, followed closely by China at 649.1, Switzerland at 610.7, and Russia at 601.4. The report on Russia offers a notably flattering assessment, describing the country as “a state with a high level of sovereignty and record military and economic self-sufficiency, a strong scientific and cultural base, but with limited flexibility of the civilian economy and dependence on internal administrative capital.” The report further asserts that Russia has established “a stable and high level of national sovereignty, from financial and military self-sufficiency to digital and cultural control over the internal environment,” and claims the country has become “one of the three world powers with full digital and military nuclear sovereignty.” This carefully constructed index served as a pseudo-academic tool to legitimize pro-Russian narratives while undermining the credibility of Western nations.
The operational footprint of the influence campaign extended beyond the IBI website itself, encompassing a coordinated social media presence designed to amplify the think tank’s content and reach target audiences. OpenAI’s investigation found that the majority of the generated content was in English, and the operators gave specific instructions to ChatGPT to conceal any linguistic clues that could allude to their Russian provenance. This attention to linguistic detail suggests a sophisticated understanding of how to avoid detection by both automated screening systems and human analysts. The campaign’s reach was relatively limited, with Telegram channels attracting the most users and amassing approximately 10,000 to 20,000 followers each. However, the operation also extended to other platforms, including Substack, X, Facebook, and LinkedIn, where AI-generated posts and comments were used to promote the IBI and its content. In one notable instance, a ChatGPT user utilized the tool to create a profile picture for a Telegram channel named “American Observer” and repeatedly requested Russian-language summaries of the channel’s activity. This cross-platform approach, combined with the use of VPNs to mask geographic origins, demonstrates a methodical effort to build a persistent presence across the information ecosystem while maintaining operational security.
OpenAI’s analysis of the operation revealed that its significance lies less in the audience it reached than in the infrastructure it had built. The company emphasized that while the actors only used ChatGPT to produce isolated promotional posts, those posts pointed to an otherwise credible-appearing institution complete with purported experts, republished academic work, and a purported proprietary risk index. This infrastructure represents a new model of influence operations, one that leverages AI not merely to generate content at scale, but to construct entire ecosystems of fake legitimacy. The operation is distinct from previously documented Russia-linked influence operations, which have typically relied on more overt tactics such as bot networks and coordinated inauthentic behavior. Instead, this campaign focused on building a durable institutional facade that could be activated and scaled over time. OpenAI noted that the supporting use of AI in the operation ultimately led to its exposure, as the company’s investigation into AI-generated social media posts uncovered the broader network. This highlights both the risks and the vulnerabilities inherent in using AI tools for covert influence operations, as the very technology that enables sophisticated disinformation campaigns also creates new opportunities for detection and attribution.
The takedown of this operation carries significant implications for the future of information warfare and the role of artificial intelligence in shaping public discourse. As AI tools become increasingly sophisticated and accessible, state-aligned actors and malicious non-state groups are likely to adopt similar strategies, using generative AI to create fake institutions, manufacture expert consensus, and produce persuasive content at scale. The IBI operation demonstrates that even relatively small-scale efforts can establish credible-appearing infrastructure that could potentially be expanded over time, making early detection and disruption critical. OpenAI’s proactive investigation and disclosure of this operation represents a growing trend among AI companies to monitor for misuse of their platforms and share threat intelligence with the broader security community. However, the company’s findings also underscore the challenges facing platform companies, researchers, and policymakers in identifying and countering AI-enabled influence operations. The use of VPNs to bypass geographic restrictions, the careful concealment of linguistic traces, and the construction of elaborate fake institutions all point to an evolving threat landscape in which traditional indicators of inauthentic behavior may no longer be sufficient. As the line between genuine and manufactured online content continues to blur, the need for robust detection mechanisms, cross-platform collaboration, and public awareness becomes ever more urgent. The OpenAI investigation serves as both a warning and a demonstration of what is possible when AI companies actively work to expose the misuse of their technologies, but it also raises questions about the broader ecosystem of AI tools and platforms that may not have equivalent safeguards in place.
