Why it matters. The operation does not appear to have reached a large audience. What makes it noteworthy is the infrastructure behind it. Rather than simply generating propaganda posts, its operators sought to manufacture the appearance of intellectual authority around an ostensibly independent research institute.
What happened. OpenAI said on Aug. 25 that it had banned a cluster of ChatGPT accounts that “very likely originated in Russia” and were being used to promote the International Burke Institute, or IBI, a self-described expert community claiming to be based in Israel.
- The investigation began with AI-generated social media content but eventually uncovered a broader influence operation involving a website, purported international experts, republished academic material and a proprietary index measuring national “sovereignty.”
- OpenAI attributed the accounts to operators very likely based in Russia. It did not attribute the campaign directly to the Russian government or intelligence services.
How it worked. The operators mostly prompted ChatGPT in Russian while asking the model to produce English-language posts for X, Facebook, LinkedIn, Substack and Telegram. They also instructed the model to remove linguistic clues that might reveal their Russian origin. Because OpenAI does not provide access to its models from Russia, the users connected through VPNs, according to the company. AI, however, was not used to build the entire operation.
- Its main role was to generate promotional posts and comments directing users toward material published by the International Burke Institute. The website itself was the more elaborate part of the scheme.
- A think tank built for credibility: The IBI website, registered in February 2025, presented the organization as an Israel-based “expert community” featuring prominent international scholars and researchers.
- Of a sample of 36 articles published between September 2025 and May 2026 and linked to experts listed on the website, 34 had been copied from elsewhere online. Some were several years old. Others were attributed to people who had not written them.
- One article on the China-Pakistan Economic Corridor, for instance, appeared to have been copied from a Cambridge University Press publication but reassigned to another academic. Another piece on migration governance was taken from the Migration Policy Institute and associated with the profile of an Australian food chemistry professor.
- The apparent objective was to take authentic academic material and use it to give the institution an air of legitimacy while obscuring where that material actually came from.
The “sovereignty” narrative. At the heart of the operation was what the institute called the Burke Index, a framework purportedly designed to assess countries’ political, economic, technological, informational, cultural and military sovereignty.
- Those reports consistently pushed a recognisable geopolitical narrative.
- Russia tended to fare well. Western states — including the United States, France, Germany and members of the European Union — were portrayed as constrained by alliances, economic dependencies or outside influence.
- OpenAI described this as the first Russia-linked influence operation it had disrupted that went to such elaborate lengths to build an ostensibly credible institution and its own “sovereignty index.”
The Italy angle. Italy appears directly in that ecosystem. One IBI report compared Italy with Slovakia under the title “Slovakia vs Italy: The paradoxes of the Sovereignty of the Hostage-Countries of the ‘Big’ Union.” OpenAI specifically cited the report as an example of the material produced around the institute’s sovereignty framework.
- The underlying narrative portrayed Italy as formally sovereign but constrained by its membership in the European Union, the euro and NATO.
- These are not invented political fault lines. Questions about national sovereignty, Brussels, relations with Washington and the costs of European and Atlantic commitments already exist inside Italy’s domestic debate.
- That makes them useful material for influence operations. Rather than introducing entirely new arguments, such campaigns can amplify existing disagreements, reorganise them around a strategic narrative and disguise their origin behind sources that appear independent.
Beyond Italy. OpenAI identified German-language content posted to a Telegram channel called Lahme Ente, which regularly criticized Ukraine, the EU and the German government while advocating better relations with Russia.
- Another operator generated logos for roughly a dozen Telegram channels focused on Germany, the United States, France, Poland and Türkiye, and repeatedly requested Russian-language summaries of their activity.
- The pattern suggests an attempt to adapt similar narratives to different national information environments rather than push a single message indiscriminately across Europe.
Reality check. The campaign was not especially successful, at least in measurable audience terms. OpenAI said most social media posts received few views and official IBI accounts had limited followings. Some Telegram channels performed better, generally attracting between 10,000 and 20,000 subscribers.
- Using the Brookings Institution’s Breakout Scale for influence operations, OpenAI placed the campaign at the lower end of Category Three – meaning it operated across several platforms and showed some evidence of reaching authentic audiences, but did not achieve a major breakthrough.
The bigger picture. The case offers a useful indication of where AI-enabled influence operations may be heading. The technology did not replace traditional information warfare, but supported it.
- The more consequential investment was in creating something that looked credible: an institution, named experts, academic publications and a methodology able to lend apparently neutral authority to politically useful conclusions.
- AI then made it cheaper to promote those assets, adapt messaging across languages and platforms, generate comments and maintain multiple online identities.
- The main risk is not simply that generative AI can produce propaganda faster. It is that it can become one layer in a much broader attempt to manufacture legitimacy around a hidden source.
- In this case, that same use of AI also became a vulnerability: activity on ChatGPT helped investigators uncover an operation extending well beyond the individual posts the model had generated.
