OpenAI Blocks Russian Influence Campaign Masquerading as Israeli Research Institute

OpenAI revealed that it has blocked a covert Russian influence operation promoting the International Burke Institute (IBI), an entity falsely presenting itself as an Israeli-based expert community. The campaign involved AI-generated social media posts, plagiarized academic articles, and a "sovereignty index" portraying Russia positively, while attempting to conceal its Russian origins. This is the first time OpenAI has disrupted such a complex Russia-linked influence effort.

The investigation began with AI-generated posts on platforms including X, LinkedIn, Facebook, Substack, and Telegram. Operators used ChatGPT, accessed via VPN to bypass regional restrictions, to create posts and responses mainly in English, despite Russian prompts, and to mask linguistic signs of Russian origin. The IBI website, registered in February 2025 and claiming Israeli roots, published numerous articles, most of which were plagiarized or machine-translated from Slavic languages, often misattributed to prominent academics such as Francis Fukuyama and Noam Chomsky.

OpenAI's analysis of 36 sampled articles found 34 were copied from other sources, including Cambridge University Press and the Migration Policy Institute, with false attributions to unrelated scholars. The IBI also published reports comparing countries’ "sovereignty" levels, often criticizing nations opposing Russia’s war in Ukraine, such as France, Germany, and the EU, while promoting narratives favorable to Russia and the US.

Additional campaign activities included German-language Telegram channels like Lahme Ente, which regularly criticized Ukraine, the EU, and Germany, advocating for improved Russia relations. Another operator created logos for a dozen Telegram channels targeting countries including Germany, the US, France, Poland, and Turkey, some promoting IBI content. One US-focused channel, American Observer, impersonated a US media outlet but contained unnatural English.

Although much content was AI-generated or plagiarized, OpenAI found indications that some real individuals in Israel may have represented IBI at conferences and in article submissions, though their connection to the Russian operators remains unclear. The campaign’s reach was limited, with official IBI accounts having few followers, while Telegram channels attracted 10,000 to 20,000 subscribers each. OpenAI categorized the operation as low-impact but highlighted its sophisticated infrastructure designed to build credibility and obscure origins over time.

OpenAI emphasized this case illustrates how AI can be leveraged to support broader influence efforts by creating authoritative assets, hiding narrative sources, and sustaining long-term disinformation campaigns.