OpenAI Bans Russian ChatGPT Accounts Used to Run Influence Operation

Summary

OpenAI has banned a cluster of Russian accounts that used VPNs to circumvent platform restrictions and employed ChatGPT to generate social media posts and comments. The material was distributed across Substack, Telegram, X, Facebook and LinkedIn as part of a coordinated influence effort. The operation centered on promoting the International Burke Institute, a fabricated entity presented as an Israel-based community of experts in economics, politics, sociology and international relations.

The institute’s website, registered in February 2025, featured copied or misattributed academic articles and a custom “Sovereignty Index,” also called the Burke Index. This metric combined seven indicators—political, economic, technological, information, cultural, cognitive and military sovereignty—into a cumulative score that placed Russia fourth globally, behind only the United States, China and Switzerland. The index relied on selective national data and expert assessments that consistently portrayed Russia as possessing high military and economic self-sufficiency while criticizing Western governance models.

Operators instructed the model to produce English-language output that concealed linguistic markers of Russian origin. Although the majority of the generated posts reached modest audiences, with Telegram channels attracting roughly 10,000 to 20,000 followers each, the activity extended to creating profile imagery and Russian-language summaries of related channels. OpenAI noted that the campaign differed from previously documented Russian influence operations primarily through its emphasis on constructing an apparently authoritative institutional facade around the index.

The case illustrates how generative tools can support the assembly of seemingly credible assets—fabricated think tanks, proprietary metrics and republished scholarship—while also leaving detectable traces that enable detection. OpenAI assessed the broader infrastructure as more significant than the immediate reach of the posts, because such assets could be expanded over time to lend weight to preferred narratives.

Why it matters

This article provides critical threat intelligence for security professionals on how state-aligned actors abuse LLMs to scale disinformation and manufacture credibility. Understanding these TTPs is essential for Dutch and EU security teams defending against AI-enabled influence operations.