The Influence Operation That Ran on Borrowed Reputations

Jessica Brandt is a leading expert on the national security implications of AI, foreign malign influence, and the governance of emerging technologies. She previously served as director of the Foreign Malign Influence Center at the Office of the Director of National Intelligence.

The International Burke Institute (IBI) presents itself as an Israel-based expert community. Its website features research papers, a proprietary sovereignty index, and a roster of affiliates that includes at least three current and former Council on Foreign Relations (CFR) experts. (None of them had ever heard of IBI.) Much of the academic work on the site appears to be copied and misattributed.

IBI appears to sit at the center of a Russian influence operation that OpenAI exposed last week. According to the firm, a group of influence actors in Russia used virtual private networks (VPNs) to log in to ChatGPT. They asked it to generate social media posts, mostly in English, that later appeared across the Western web, on platforms including X, LinkedIn, Facebook, Substack, and Telegram. The actors instructed ChatGPT to hide any linguistic clues that they were Russian.

The posts tended to criticize Ukraine, the European Union, and other Western governments, and they advocated for better relations with Russia. They also heavily promoted IBI and its sovereignty index, which praises Russia and denigrates the West.

IBI pushed back in a heated statement, disputing reporting by Le Monde on the OpenAI report. It did not, however, address the fact that UNICEF denied IBI was a partner institution—contrary to what IBI’s website still states. A person’s name is not attached to the statement. IBI does not list a founder or leader. The website remains live and it continues to post content.  

This influence operation’s reach was modest. Most posts did not receive a substantial number of views. OpenAI placed this operation at the low end of level three of the Breakout Scale, a six-level framework that researchers use to characterize the reach of influence operations. IBI’s level is characterized as reaching multiple platforms, with some indications of breakout to authentic audiences.

While its reach isn’t particularly notable, the operation’s creation of substantial infrastructure certainly is. This effort built a credible-appearing institution complete with purported expert affiliates, research papers, and a proprietary sovereignty index. Its perpetrators used AI as a tool to promote that infrastructure in multiple languages, without the language mistakes that previously gave campaigns away, and often using AI-generated personas, profile photos, and face-to-camera video reels. The campaign thus highlights the ways that AI is enabling influence actors to produce targeted content at scale and complicate attribution—accelerating, if not wholly revolutionizing, foreign information operations.

The operators even attempted to seed fabricated research papers featuring fictitious authors on authentic academic research-hosting platforms, according to Meta, which disrupted related activity on Facebook and Instagram after reviewing information shared by OpenAI. Russia has long attempted to launder campaigns across multiple platforms and to manufacture credibility by disguising their content as the product of authentic domestic voices. These attempts to work across the full information environment and better hide their hand while doing so are emblematic of the maturation of influence campaigns over the past decade.

Broadly, the episode highlights the value of regular threat intelligence reporting from AI labs, which have unique insight into consequential activity that social media platforms and governments may not. Over the past decade, social media companies have routinized their threat intelligence reporting—AI labs should too. 

Consistent industry reporting would create an important public-interest record of foreign influence activity, offer investigative leads for academic researchers and journalists seeking to build a picture of this activity as it evolves, and enable other companies to act as appropriate. This alone would help policymakers in and out of government deal with current threats and anticipate those to come.