OpenAI Disrupts Russian Influence Ops

OpenAI has disrupted a new covert influence operation originating from Russia, which utilized its own AI models to generate deceptive content. The campaign focused on promoting a self-described Israeli think tank, the International Burke Institute (IBI), and aimed to cast Russia in a favorable light while denigrating Western countries. This detailed operation, uncovered through investigations into AI-generated social media posts, reveals an elaborate effort to manipulate public opinion.

AI as a Tool for Deception

The Russian operators accessed OpenAI's models using VPNs, prompting in Russian to create social media comments in English. They specifically instructed ChatGPT to mask any linguistic indicators of their Russian origin. These generated comments were disseminated across platforms including Substack, Telegram, X (formerly Twitter), Facebook, and LinkedIn, often promoting content from the IBI website.

The International Burke Institute Facade

The IBI website, registered in February 2025, presented itself as an expert community based in Israel. However, OpenAI's analysis found that many articles published on the site were plagiarized or misattributed academic works. Some content appeared to be machine-translated from a Slavic language, evidenced by awkward phrasing like "the Svetofor coalition" to describe Germany's "traffic-light coalition," where "svetofor" is a Slavic word for traffic light.

Sophisticated Infrastructure and Tactics

Beyond social media posts, the operation developed a "sovereignty index" designed to praise Russia and criticize nations like France, Germany, and the United States. The IBI website also featured copied academic articles, some misattributed to prominent academics like Francis Fukuyama and Noam Chomsky to lend an air of legitimacy. This elaborate construction distinguishes it from simpler influence campaigns previously disrupted.

Broader AI and Influence Operation Context

The campaign's use of AI was primarily for generating promotional social media content and replies to real users on platforms like Substack. Operators also created logos for various Telegram channels focused on Germany, the USA, France, Poland, and Türkiye, some of which amplified IBI's narratives. One operator even generated profile pictures for these channels, exhibiting non-idiomatic English in bios such as "a totally unhackneyed perspective on hazzy." StartupHub.ai data indicates that while platforms like Substack have a low score of 10/100, Telegram scores a more substantial 57/100 in terms of user engagement and reach within the current digital media landscape.

Impact and Implications

While the immediate audience reach of this operation appears limited, with social media posts garnering low views, the established infrastructure is significant. The campaign's strength lay in its creation of a seemingly credible institution, complete with purported experts and research, to obscure the true origins of favored narratives. OpenAI assessed the operation at the lower end of Category Three on the Brookings Breakout Scale, indicating multiple platform uses with some breakout to authentic audiences. This incident underscores how influence actors can use AI as a supporting tool to manufacture authority and establish assets that could be scaled over time.

AI's Dual Nature in Information Warfare

The operation's reliance on AI also led to its exposure, as the generated content and linguistic anomalies provided investigative clues. This case serves as a potent reminder of AI's dual-use potential in information warfare, enabling both the detection and execution of sophisticated influence campaigns. The elaborate nature of this Russian operation, as detailed in OpenAI's findings, signifies an evolving threat landscape where AI-generated content is increasingly integrated into state-sponsored disinformation efforts.